Comments on AI and math
While this is not about categories specifically I have some comments I would like to make public and this would seem to be my outlet. So please bear with me. The main thing I want to say about math and AI is that AI appears to be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career. I spent a month in Fribourg in 1997 and during that month, Heinrich Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper. During the year 1969-70, my second at McGill, I was working on regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so. So, I took an unpaid leave and headed to Fribourg, On the way there, I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category. So two questions, one from Peter and one from Heinrich, had elevated what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers. Michael
thank you for interesting examples, mike :) while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos. a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols. the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that. if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers... i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities... 2c, -- dusko On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Well, if the LLMs owners offer a large share of their profit with us, I am sure more people will be interested in sharing "question-generation processes". On Mon, Nov 9, 2026 at 11:03 AM Dusko Pavlovic via Categories < categories-list@categories.org.au> wrote:
thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments I
would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be
excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular
and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I
stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
-- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 Cellphone: +27(0)71 124 5014 Homepage: https://sites.google.com/uwc.ac.za/ando/home www.uwc.ac.za <https://www.google.com/url?q=https://uwc.ac.za&source=gmail-html&ust=1655289284799000&usg=AOvVaw2MXM9-xWRLqwk62e6juGPq> -- Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer <https://www.uwc.ac.za/disclaimer>
Now that the CT focus is oriented towards 'questions', if I may, I thought my good friend's "The Mathematics of Questions": <https://arxiv.org/pdf/2503.18992> might have some needed infrastructure to judge a question as such (on a lighter note, in my Godavari river region: Character One: what is your name Character Two: what is NOT my name ;) as part of the invariant work-horse tugging across the three avatars of AI: PRE-processing. Happy Tuesday :) Thanking you, Yours respectfully, posina ---- On Mon, 09 Nov 2026 16:37:03 +0530 categories-list@categories.org.au wrote ---- Well, if the LLMs owners offer a large share of their profit with us, I am sure more people will be interested in sharing "question-generation processes". On Mon, Nov 9, 2026 at 11:03 AM Dusko Pavlovic via Categories < mailto:categories-list@categories.org.au > wrote: thank you for interesting examples, mike :) while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos. a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols. the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that. if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers... i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities... 2c, -- dusko On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories < mailto:categories-list@categories.org.au > wrote:
While this is not about categories specifically I have some comments I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________
Categories mailing list -- mailto:categories-list@categories.org.au
To unsubscribe send an email to mailto:categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- mailto:categories-list@categories.org.au To unsubscribe send an email to mailto:categories-list-leave@categories.org.au -- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 Cellphone: +27(0)71 124 5014 Homepage: https://sites.google.com/uwc.ac.za/ando/home https://www.google.com/url?q=https://uwc.ac.za&source=gmail-html&ust=1655289284799000&usg=AOvVaw2MXM9-xWRLqwk62e6juGPq Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer _______________________________________________ Categories mailing list -- mailto:categories-list@categories.org.au To unsubscribe send an email to mailto:categories-list-leave@categories.org.au
Marx, with his 'tools as extended body parts': https://www.nature.com/articles/s41586-018-0460-0#article-comments would be happy reading your: "we should consider models of our thinking processes as a part of ourselves". Thanking you, Yours respectfully, posina On Mon, 9 Nov, 2026, 2:40 pm Dusko Pavlovic via Categories, < categories-list@categories.org.au> wrote:
thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments I
would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be
excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular
and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I
stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
To extract a citation from the reference: *It is precisely because earth and tools are man's extended body that property, which severs the worker from his extended body, places him in the position of selling not his work but his now-truncated activity of labor: thus the idea is central to Capital" (Scarry, 1994, p. 85).* On Tue, Nov 10, 2026 at 12:09 PM Posina Venkata Rayudu via Categories < categories-list@categories.org.au> wrote:
Marx, with his 'tools as extended body parts':
https://www.nature.com/articles/s41586-018-0460-0#article-comments
would be happy reading your:
"we should consider models of our thinking processes as a part of ourselves".
Thanking you, Yours respectfully, posina
On Mon, 9 Nov, 2026, 2:40 pm Dusko Pavlovic via Categories, < categories-list@categories.org.au> wrote:
thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments I
would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be
excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular
and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I
stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
-- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 Cellphone: +27(0)71 124 5014 Homepage: https://sites.google.com/uwc.ac.za/ando/home www.uwc.ac.za <https://www.google.com/url?q=https://uwc.ac.za&source=gmail-html&ust=1655289284799000&usg=AOvVaw2MXM9-xWRLqwk62e6juGPq> -- Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer <https://www.uwc.ac.za/disclaimer>
why is someone who bundles marx with feuerbach (apparently unaware of the 10 theses Ueber Feuerbach) a suitable reference for the "the idea central to Capital"? she is a professor of english writing her books for her audience, and can afford a bird's eye view of marx. but what pretext do we have to replace marx by a bird's eye view of her bird's eye view? On Tue, Nov 10, 2026 at 12:24 AM Ando Razafindrakoto < arazafindrakoto@uwc.ac.za> wrote:
To extract a citation from the reference: *It is precisely because earth and tools are man's extended body that property, which severs the worker from his extended body, places him in the position of selling not his work but his now-truncated activity of labor: thus the idea is central to Capital" (Scarry, 1994, p. 85).*
On Tue, Nov 10, 2026 at 12:09 PM Posina Venkata Rayudu via Categories < categories-list@categories.org.au> wrote:
Marx, with his 'tools as extended body parts':
https://www.nature.com/articles/s41586-018-0460-0#article-comments
would be happy reading your:
"we should consider models of our thinking processes as a part of ourselves".
Thanking you, Yours respectfully, posina
On Mon, 9 Nov, 2026, 2:40 pm Dusko Pavlovic via Categories, < categories-list@categories.org.au> wrote:
thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments I
would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to
be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular
and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I
stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to
categories-list-leave@categories.org.au _______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
-- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 <+27%2021%20959%203018> Cellphone: +27(0)71 124 5014 <+27%2071%20124%205014> Homepage: https://sites.google.com/uwc.ac.za/ando/home www.uwc.ac.za <https://www.google.com/url?q=https://uwc.ac.za&source=gmail-html&ust=1655289284799000&usg=AOvVaw2MXM9-xWRLqwk62e6juGPq>
Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer
Would you want a quote from a reference in Lawvere's paper "Quantifiers and Sheaves" which concerns itself with the origin of correct ideas? On Thu, Nov 12, 2026 at 8:50 PM Dusko Pavlovic <duskgoo@gmail.com> wrote:
why is someone who bundles marx with feuerbach (apparently unaware of the 10 theses Ueber Feuerbach) a suitable reference for the "the idea central to Capital"? she is a professor of english writing her books for her audience, and can afford a bird's eye view of marx. but what pretext do we have to replace marx by a bird's eye view of her bird's eye view?
On Tue, Nov 10, 2026 at 12:24 AM Ando Razafindrakoto < arazafindrakoto@uwc.ac.za> wrote:
To extract a citation from the reference: *It is precisely because earth and tools are man's extended body that property, which severs the worker from his extended body, places him in the position of selling not his work but his now-truncated activity of labor: thus the idea is central to Capital" (Scarry, 1994, p. 85).*
On Tue, Nov 10, 2026 at 12:09 PM Posina Venkata Rayudu via Categories < categories-list@categories.org.au> wrote:
Marx, with his 'tools as extended body parts':
https://www.nature.com/articles/s41586-018-0460-0#article-comments
would be happy reading your:
"we should consider models of our thinking processes as a part of ourselves".
Thanking you, Yours respectfully, posina
On Mon, 9 Nov, 2026, 2:40 pm Dusko Pavlovic via Categories, < categories-list@categories.org.au> wrote:
thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments
I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to
be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on
regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there,
I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to
categories-list-leave@categories.org.au _______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
-- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 <+27%2021%20959%203018> Cellphone: +27(0)71 124 5014 <+27%2071%20124%205014> Homepage: https://sites.google.com/uwc.ac.za/ando/home www.uwc.ac.za <https://www.google.com/url?q=https://uwc.ac.za&source=gmail-html&ust=1655289284799000&usg=AOvVaw2MXM9-xWRLqwk62e6juGPq>
Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer
-- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 Cellphone: +27(0)71 124 5014 Homepage: https://sites.google.com/uwc.ac.za/ando/home www.uwc.ac.za <https://www.google.com/url?q=https://uwc.ac.za&source=gmail-html&ust=1655289284799000&usg=AOvVaw2MXM9-xWRLqwk62e6juGPq> -- Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer <https://www.uwc.ac.za/disclaimer>
lawvere certainly didn't advise anyone to study philosophy through quotations, authoritative or not. if you try reading marx, you will understand why many readers have many readings. marx of all people shouldn't be reduced to interpretations, knowing that so many people used their interpretations to justify their crimes. (incidentally, it is now echoing in my head that i wrote "10 theses Ueber Feuerbach", didn't i? there were actually 11 theses. maybe my aging brain mixed up marx and moses 🤣 speaking of misinterpretations :)))
I read Marx's capital which starts with the accumulation of capital from the extraction of labour (including so-called "skilled labour", which is discussed briefly in Chapter 1 as well.) And that was the whole point of the initial citation. Marx is not only Feuerbach. Marx is Feuerbach (materialism) and Hegel (dialectics). On Thu, Nov 12, 2026 at 10:16 PM Dusko Pavlovic <duskgoo@gmail.com> wrote:
lawvere certainly didn't advise anyone to study philosophy through quotations, authoritative or not. if you try reading marx, you will understand why many readers have many readings. marx of all people shouldn't be reduced to interpretations, knowing that so many people used their interpretations to justify their crimes.
(incidentally, it is now echoing in my head that i wrote "10 theses Ueber Feuerbach", didn't i? there were actually 11 theses. maybe my aging brain mixed up marx and moses 🤣 speaking of misinterpretations :)))
-- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 Cellphone: +27(0)71 124 5014 Homepage: https://sites.google.com/uwc.ac.za/ando/home www.uwc.ac.za <https://www.google.com/url?q=https://uwc.ac.za&source=gmail-html&ust=1655289284799000&usg=AOvVaw2MXM9-xWRLqwk62e6juGPq> -- Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer <https://www.uwc.ac.za/disclaimer>
With due apologies to the creative types, I remember reading "creativity", which made me hasten to share a more democratic reading of, than is likely to be the case with the aforementioned, creativity: Nudging AI towards Truth https://www.nature.com/articles/s41586-024-07930-y#article-comments Since we are implicitly, if not explicitly, attending to the contrast: procedural knowledge vs. declarative understanding it seems to me that a test of understanding (in contrast to mere knowing) is, if not 'write in your own words' (https://playinmath.wordpress.com/2023/02/08/history-and-philosophy-of-mathem...), a more detailed instruction, a translation of the given conceptual repertoire into a desired one (of the audience), along with the corresponding presentations (words / sentences / displays) of the given and the desired repertoires of abstract generals. Going by Feynman's experience of failing to explain to a non-colleague at the MIT math dept., a proxy for understanding can be in terms of the category of map objects of functors between categories as abstract generals. Linguistic proficiency in a large number of languages (one of our prime ministers knew some 24 languages) is 'good for' communicating with diverse audience, while proficiency in concepts, their abstraction, interpretations, and laws of rational passage between concepts beyond all that logic that is familiar in going from a given sentence to a sentence as its conclusion (https://philpapers.org/rec/POSLOC) is 'good for' bringing into sharp focus the reality that is reflected in our thinking / experience. In closing, getting the matter that moves to move in directions desired by a thinking matter is beyond the reach of moving matter, and is what distinguishes matters that think: make falling matter fly :). Surely, there is something about AI---AI & I---conversation that's radiating tomorrow worth celebrating today (for those panic, there's petril ;) the deep history of thinking about things and about making things thought in the privacy of one's mind in conversation, a pleasant conversation (I have to tell all AIs to stop flattering for flattery and fame are not conducive to proper thinking which I value.), a fruitful conversation, the long-lost habit of the much-desired iron-manning of a meeting of matters---moving and thinking---material reflections of reality subsuming the architecture of understanding, a case study of intellect to document, a Bohr atom of understanding (instead of an abstract approach of, say, calculating the truth value triangle of a category of syllogisms <https://philpapers.org/rec/POSSAL>). On Fri, Nov 13, 2026 at 1:55 AM Ando Razafindrakoto <arazafindrakoto@uwc.ac.za> wrote:
I read Marx's capital which starts with the accumulation of capital from the extraction of labour (including so-called "skilled labour", which is discussed briefly in Chapter 1 as well.) And that was the whole point of the initial citation. Marx is not only Feuerbach. Marx is Feuerbach (materialism) and Hegel (dialectics).
On Thu, Nov 12, 2026 at 10:16 PM Dusko Pavlovic <duskgoo@gmail.com> wrote:
lawvere certainly didn't advise anyone to study philosophy through quotations, authoritative or not. if you try reading marx, you will understand why many readers have many readings. marx of all people shouldn't be reduced to interpretations, knowing that so many people used their interpretations to justify their crimes.
(incidentally, it is now echoing in my head that i wrote "10 theses Ueber Feuerbach", didn't i? there were actually 11 theses. maybe my aging brain mixed up marx and moses 🤣 speaking of misinterpretations :)))
-- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 Cellphone: +27(0)71 124 5014 Homepage: https://sites.google.com/uwc.ac.za/ando/home www.uwc.ac.za
Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer
On Understanding and Creativity Modern sub-symbolic (neural network-based) AI has brought renewed focus to the concepts of "understanding" and "explaining" automated decision-making processes (e.g., the XAI initiative). Many have begun to ask: what, exactly, is understanding? Spinoza once remarked—to paraphrase from memory—that we understand something when we are able to reconstruct (build) it within our own minds. It occurred to me that this process of construction can be reduced to the principles of conservation. Group theory and category theory characterize these conservation laws, which, in turn, boil down to constructing the object of our inquiry from a foundational set of elements—referred to as "generators" in group theory and "basic objects" in category theory. Both theories embody the same philosophical and scientific principle: to understand a whole is to identify its minimal generating basis and to describe the transformations that leave this basis invariant. Both approaches address a fundamental challenge: how to describe an infinite or colossally complex whole using a finite set of rules. Group theory achieves this through the internal symmetry of objects, while category theory does so through the symmetry of the relationships between them. In the creative work of a mathematician, an aesthetic sense of beauty plays a significant role. And beauty, at its core, is symmetry. It seems likely that, in time, AI will also attain a formalized understanding of beauty—by way of group theory and category theory. пт, 13 нояб. 2026 г. в 23:12, Posina Venkata Rayudu via Categories < categories-list@categories.org.au>:
With due apologies to the creative types, I remember reading "creativity", which made me hasten to share a more democratic reading of, than is likely to be the case with the aforementioned, creativity:
Nudging AI towards Truth https://www.nature.com/articles/s41586-024-07930-y#article-comments
Since we are implicitly, if not explicitly, attending to the contrast:
procedural knowledge vs. declarative understanding
it seems to me that a test of understanding (in contrast to mere knowing) is, if not 'write in your own words' ( https://playinmath.wordpress.com/2023/02/08/history-and-philosophy-of-mathem... ), a more detailed instruction, a translation of the given conceptual repertoire into a desired one (of the audience), along with the corresponding presentations (words / sentences / displays) of the given and the desired repertoires of abstract generals. Going by Feynman's experience of failing to explain to a non-colleague at the MIT math dept., a proxy for understanding can be in terms of the category of map objects of functors between categories as abstract generals. Linguistic proficiency in a large number of languages (one of our prime ministers knew some 24 languages) is 'good for' communicating with diverse audience, while proficiency in concepts, their abstraction, interpretations, and laws of rational passage between concepts beyond all that logic that is familiar in going from a given sentence to a sentence as its conclusion (https://philpapers.org/rec/POSLOC) is 'good for' bringing into sharp focus the reality that is reflected in our thinking / experience. In closing, getting the matter that moves to move in directions desired by a thinking matter is beyond the reach of moving matter, and is what distinguishes matters that think: make falling matter fly :). Surely, there is something about AI---AI & I---conversation that's radiating tomorrow worth celebrating today (for those panic, there's petril ;) the deep history of thinking about things and about making things thought in the privacy of one's mind in conversation, a pleasant conversation (I have to tell all AIs to stop flattering for flattery and fame are not conducive to proper thinking which I value.), a fruitful conversation, the long-lost habit of the much-desired iron-manning of a meeting of matters---moving and thinking---material reflections of reality subsuming the architecture of understanding, a case study of intellect to document, a Bohr atom of understanding (instead of an abstract approach of, say, calculating the truth value triangle of a category of syllogisms <https://philpapers.org/rec/POSSAL>).
On Fri, Nov 13, 2026 at 1:55 AM Ando Razafindrakoto <arazafindrakoto@uwc.ac.za> wrote:
I read Marx's capital which starts with the accumulation of capital from
Marx is not only Feuerbach. Marx is Feuerbach (materialism) and Hegel (dialectics).
On Thu, Nov 12, 2026 at 10:16 PM Dusko Pavlovic <duskgoo@gmail.com> wrote:
lawvere certainly didn't advise anyone to study philosophy through
quotations, authoritative or not. if you try reading marx, you will understand why many readers have many readings. marx of all people shouldn't be reduced to interpretations, knowing that so many people used
the extraction of labour (including so-called "skilled labour", which is discussed briefly in Chapter 1 as well.) And that was the whole point of the initial citation. their interpretations to justify their crimes.
(incidentally, it is now echoing in my head that i wrote "10 theses
Ueber Feuerbach", didn't i? there were actually 11 theses. maybe my aging brain mixed up marx and moses 🤣 speaking of misinterpretations :)))
-- Ando Razafindrakoto Senior Lecturer Faculty of Natural Sciences Department of Mathematics and Applied Mathematics Landline: +27(0)21 959 3018 <+27%2021%20959%203018> Cellphone: +27(0)71 124 5014 <+27%2071%20124%205014> Homepage: https://sites.google.com/uwc.ac.za/ando/home www.uwc.ac.za
Disclaimer - This e-mail is subject to UWC policies and e-mail disclaimer published on our website at: https://www.uwc.ac.za/disclaimer
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
I've given some thought to this and I just don't see what AI is going to train on. For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly. For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results. As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that. Michael ________________________________ From: Dusko Pavlovic via Categories <categories-list@categories.org.au> Sent: Monday, November 9, 2026 3:58 AM To: Michael Barr, Prof. via Categories <categories-list@categories.org.au> Cc: Dusko Pavlovic <duskgoo@gmail.com> Subject: [categories] Re: Comments on AI and math thank you for interesting examples, mike :) while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos. a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols. the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that. if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers... i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities... 2c, -- dusko On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Dear Michael, I concur with your thoughts and it goes back to the fundamental question: what is creativity, what is the process that leads to new ideas and is it something that machines can do? The current conversation around AI is really about LLMs and their interaction as "agents". As such, these are good at absorbing vast amounts of human-produced (and now fed-back machine-produced) writing. However, this comes with the caveat that "new ideas" are applications of patterns that the model has been trained on. Thus, is creativity then mere pattern-recognition or is there more to it? I think that many of us want to believe that there is, but this will not prevent any profit-oriented company from saying otherwise (thank you Ando for rectifying the use of Marx here!). This leads to a, may I say, spiritual question that is lurking here: is maths a game that can be won by machines that solve problems with no end, and may even ask questions with no end, at a pace that no human can keep up with so that maths should no longer be a human endeavour? There are efforts of many mathematicians (and those that essentially do maths in other fields) to address the issues arising from LLMs being capable of solving maths problems (https://leidendeclaration.ai/, https://mathandai.org/). However, just as (financial) capitalism is a system set on self-preservation, no matter the costs, A"I" will be reinforced by companies pushing for it and researchers forced to by competition for funding and position. I see difficult times for our fields and the people in it ahead, let alone the general societal and environmental repercussions and damage this has and will only increasingly have. My apologies for this sceptical reply, but hope that we are not blindly endorsing a technological solution to the academic issues that we have and keep technology what it is: a tool to solve a particular problem. Best, Henning PS: I am reminded of mathematics as a mental process that was used by the intuitionist school in order to develop things like Kripke semantics and the philosophy in general. I wonder what the people from that time would have to say. PPS: For those not on the types mailing list, our esteemed colleague James McKinna made the good point that this discussion cannot be separated from the political questions and suggested some nice reading material: * Simon Schaffer's speech <https://erasmusprijs.org/en/laureates/simon-schaffer/acceptance-speech/> * and U. Chicago Press "Working Knowledge: A Simon Schaffer Reader" * Randy Pollack's "How to believe a machine-checked proof" On 10/11/2026 21:21, Michael Barr, Prof. via Categories wrote:
I've given some thought to this and I just don't see what AI is going to train on.
For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42^nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly.
For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results.
As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that.
Michael ------------------------------------------------------------------------ *From:* Dusko Pavlovic via Categories <categories-list@categories.org.au> *Sent:* Monday, November 9, 2026 3:58 AM *To:* Michael Barr, Prof. via Categories <categories-list@categories.org.au> *Cc:* Dusko Pavlovic <duskgoo@gmail.com> *Subject:* [categories] Re: Comments on AI and math thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments
I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to
be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on
regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there,
I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Dear All, Recently, Professor Emily Riehl published her Expert Voice in the highly reputed Science ("AI has solved many math problems, but it has not solved math," <https://doi.org/10.1126/science.ael8339>), which rightly dampens the irrational exuberance attendant automated needle-hunting. And yet the collective anxiety sweeping academia---mathematics in particular and sciences in general---exposes a deeper, internal pathology. Mathematics is at far greater risk from its own than from any external machine (cf. making vs. creating < https://www.nature.com/articles/s41586-024-07522-w#article-comments>). Our crisis is not the "virus" of AI, but is the degraded "terrain" of contemporary mathematical research, education, and applications. A self-reifying priesthood has diminished our galactic-scale space of mathematics devoted to the study of objective qualities < https://lawverearchives.com/wp-content/uploads/2025/01/2007-Axiomatic-cohesion.pdf>, objective number theories < https://conceptualmathematics.wordpress.com/wp-content/uploads/2023/05/potato.pdf>, objective logics <https://philpapers.org/rec/POSOLO>, which are intrinsic to given universes of discourse and developing mathematical methods to deal with countless integration problems < https://lawverearchives.com/wp-content/uploads/2024/12/1982-introduction-to-categories-in-continuum-physics.pdf> that are never out-of-sight in the everyday physics of cohesively extended objects populating our commonplace conscious experience < https://lawverearchives.com/wp-content/uploads/2025/01/2017.Everyday-physics-of-extended-bodies-or-why-functionals-need-analyzing.pdf>, just to name a few, into a sterile matrix of statistical significance. Mathematics is rooted, as the mathematics of, among others, Professor F. William Lawvere brought into figural salience for all to see, in the primacy of concepts—like cohesion ("stick-together") characterizing space—rather than the secondary syntax of symbolic presentations <https://lawverearchives.com/>. When this factorization is conveniently forgotten, we find ourselves mistaking a technologically sophisticated parroting for an architecture of understanding. A good antidote to this familiar form of fear-mongering lies in practicing mathematics in an inclusive-mode. We see this in the profound, quiet example of Madam Fatima Lawvere, who dedicated her life to imagining in great detail what was established as impossible, and then brought it into being: her mind-made Conceptual Mathematics textbook, thanks to which we all can learn and teach the architecture of understanding to one and all < https://categories.org.au/archives/list/categories-list@categories.org.au/me...
.
Lest the moral is lost: We do not "solve" mathematics any more than we "solve" life or music. AND We consciously participate in the practice of mathematics in aligning reason with experience. Happy Weekend :) Thanking you, Yours respectfully, posina Acknowledgement: I thank Google's Gemini for helping me organize my thoughts. On Thu, 12 Nov, 2026, 1:33 pm Henning Basold via Categories, < categories-list@categories.org.au> wrote:
Dear Michael,
I concur with your thoughts and it goes back to the fundamental question: what is creativity, what is the process that leads to new ideas and is it something that machines can do? The current conversation around AI is really about LLMs and their interaction as "agents". As such, these are good at absorbing vast amounts of human-produced (and now fed-back machine-produced) writing. However, this comes with the caveat that "new ideas" are applications of patterns that the model has been trained on. Thus, is creativity then mere pattern-recognition or is there more to it? I think that many of us want to believe that there is, but this will not prevent any profit-oriented company from saying otherwise (thank you Ando for rectifying the use of Marx here!).
This leads to a, may I say, spiritual question that is lurking here: is maths a game that can be won by machines that solve problems with no end, and may even ask questions with no end, at a pace that no human can keep up with so that maths should no longer be a human endeavour? There are efforts of many mathematicians (and those that essentially do maths in other fields) to address the issues arising from LLMs being capable of solving maths problems (https://leidendeclaration.ai/, https://mathandai.org/). However, just as (financial) capitalism is a system set on self-preservation, no matter the costs, A"I" will be reinforced by companies pushing for it and researchers forced to by competition for funding and position. I see difficult times for our fields and the people in it ahead, let alone the general societal and environmental repercussions and damage this has and will only increasingly have.
My apologies for this sceptical reply, but hope that we are not blindly endorsing a technological solution to the academic issues that we have and keep technology what it is: a tool to solve a particular problem.
Best,
Henning
PS: I am reminded of mathematics as a mental process that was used by the intuitionist school in order to develop things like Kripke semantics and the philosophy in general. I wonder what the people from that time would have to say.
PPS: For those not on the types mailing list, our esteemed colleague James McKinna made the good point that this discussion cannot be separated from the political questions and suggested some nice reading material:
* Simon Schaffer's speech <https://erasmusprijs.org/en/laureates/simon-schaffer/acceptance-speech/>
* and U. Chicago Press "Working Knowledge: A Simon Schaffer Reader"
* Randy Pollack's "How to believe a machine-checked proof"
On 10/11/2026 21:21, Michael Barr, Prof. via Categories wrote:
I've given some thought to this and I just don't see what AI is going to train on.
For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42^nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly.
For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results.
As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that.
Michael ------------------------------------------------------------------------ *From:* Dusko Pavlovic via Categories <categories-list@categories.org.au
*Sent:* Monday, November 9, 2026 3:58 AM *To:* Michael Barr, Prof. via Categories <categories-list@categories.org.au> *Cc:* Dusko Pavlovic <duskgoo@gmail.com> *Subject:* [categories] Re: Comments on AI and math thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments
I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to
be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on
regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there,
I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to
categories-list-leave@categories.org.au _______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Are you here to provoke people? People who have concerns over AI involving academic integrity, math education, accessibility are not fear-mongering. It is very in-appropriate and in very poor taste to belittle people with who have legitimate concerns. While I am sure we all respect Emily and the authors of the papers you cited, calls for authority do not establish a precedent on an issue and is a legitimate logical fallacy meant to belittle people. You should be ashamed of yourself with how you choose to look down on people and try to bully people. Perhaps (and this something you should think about) the concern over AI is not how this will affect mathematics but rather it will affect people: I am not sure you have that capacity given the message you wrote. Johnny *Johnathon Taylor* On Fri, Nov 13, 2026, 1:13 PM Posina Venkata Rayudu via Categories < categories-list@categories.org.au> wrote:
Dear All,
Recently, Professor Emily Riehl published her Expert Voice in the highly reputed Science ("AI has solved many math problems, but it has not solved math," <https://doi.org/10.1126/science.ael8339>), which rightly dampens the irrational exuberance attendant automated needle-hunting. And yet the collective anxiety sweeping academia---mathematics in particular and sciences in general---exposes a deeper, internal pathology. Mathematics is at far greater risk from its own than from any external machine (cf. making vs. creating < https://www.nature.com/articles/s41586-024-07522-w#article-comments>).
Our crisis is not the "virus" of AI, but is the degraded "terrain" of contemporary mathematical research, education, and applications. A self-reifying priesthood has diminished our galactic-scale space of mathematics devoted to the study of objective qualities < https://lawverearchives.com/wp-content/uploads/2025/01/2007-Axiomatic-cohesion.pdf>, objective number theories < https://conceptualmathematics.wordpress.com/wp-content/uploads/2023/05/potato.pdf>, objective logics <https://philpapers.org/rec/POSOLO>, which are intrinsic to given universes of discourse and developing mathematical methods to deal with countless integration problems < https://lawverearchives.com/wp-content/uploads/2024/12/1982-introduction-to-categories-in-continuum-physics.pdf> that are never out-of-sight in the everyday physics of cohesively extended objects populating our commonplace conscious experience < https://lawverearchives.com/wp-content/uploads/2025/01/2017.Everyday-physics-of-extended-bodies-or-why-functionals-need-analyzing.pdf>, just to name a few, into a sterile matrix of statistical significance.
Mathematics is rooted, as the mathematics of, among others, Professor F. William Lawvere brought into figural salience for all to see, in the primacy of concepts—like cohesion ("stick-together") characterizing space—rather than the secondary syntax of symbolic presentations <https://lawverearchives.com/>. When this factorization is conveniently forgotten, we find ourselves mistaking a technologically sophisticated parroting for an architecture of understanding.
A good antidote to this familiar form of fear-mongering lies in practicing mathematics in an inclusive-mode. We see this in the profound, quiet example of Madam Fatima Lawvere, who dedicated her life to imagining in great detail what was established as impossible, and then brought it into being: her mind-made Conceptual Mathematics textbook, thanks to which we all can learn and teach the architecture of understanding to one and all < https://categories.org.au/archives/list/categories-list@categories.org.au/me...
.
Lest the moral is lost:
We do not "solve" mathematics any more than we "solve" life or music.
AND
We consciously participate in the practice of mathematics in aligning reason with experience.
Happy Weekend :)
Thanking you, Yours respectfully, posina
Acknowledgement: I thank Google's Gemini for helping me organize my thoughts.
On Thu, 12 Nov, 2026, 1:33 pm Henning Basold via Categories, < categories-list@categories.org.au> wrote:
Dear Michael,
I concur with your thoughts and it goes back to the fundamental question: what is creativity, what is the process that leads to new ideas and is it something that machines can do? The current conversation around AI is really about LLMs and their interaction as "agents". As such, these are good at absorbing vast amounts of human-produced (and now fed-back machine-produced) writing. However, this comes with the caveat that "new ideas" are applications of patterns that the model has been trained on. Thus, is creativity then mere pattern-recognition or is there more to it? I think that many of us want to believe that there is, but this will not prevent any profit-oriented company from saying otherwise (thank you Ando for rectifying the use of Marx here!).
This leads to a, may I say, spiritual question that is lurking here: is maths a game that can be won by machines that solve problems with no end, and may even ask questions with no end, at a pace that no human can keep up with so that maths should no longer be a human endeavour? There are efforts of many mathematicians (and those that essentially do maths in other fields) to address the issues arising from LLMs being capable of solving maths problems (https://leidendeclaration.ai/, https://mathandai.org/). However, just as (financial) capitalism is a system set on self-preservation, no matter the costs, A"I" will be reinforced by companies pushing for it and researchers forced to by competition for funding and position. I see difficult times for our fields and the people in it ahead, let alone the general societal and environmental repercussions and damage this has and will only increasingly have.
My apologies for this sceptical reply, but hope that we are not blindly endorsing a technological solution to the academic issues that we have and keep technology what it is: a tool to solve a particular problem.
Best,
Henning
PS: I am reminded of mathematics as a mental process that was used by the intuitionist school in order to develop things like Kripke semantics and the philosophy in general. I wonder what the people from that time would have to say.
PPS: For those not on the types mailing list, our esteemed colleague James McKinna made the good point that this discussion cannot be separated from the political questions and suggested some nice reading material:
* Simon Schaffer's speech <https://erasmusprijs.org/en/laureates/simon-schaffer/acceptance-speech/>
* and U. Chicago Press "Working Knowledge: A Simon Schaffer Reader"
* Randy Pollack's "How to believe a machine-checked proof"
On 10/11/2026 21:21, Michael Barr, Prof. via Categories wrote:
I've given some thought to this and I just don't see what AI is going to train on.
For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42^nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly.
For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results.
As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that.
Michael ------------------------------------------------------------------------ *From:* Dusko Pavlovic via Categories < categories-list@categories.org.au> *Sent:* Monday, November 9, 2026 3:58 AM *To:* Michael Barr, Prof. via Categories <categories-list@categories.org.au> *Cc:* Dusko Pavlovic <duskgoo@gmail.com> *Subject:* [categories] Re: Comments on AI and math thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments
I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to
be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on
regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there,
I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to
categories-list-leave@categories.org.au _______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
the point that AI cannot be separated from its politics was also thoroughly researched and powerfully presented by jill lepore: https://www.jilllepore.com/the-rise-and-fall-of-the-artificial-state On Wed, Nov 11, 2026, 22:05 Henning Basold via Categories < categories-list@categories.org.au> wrote:
Dear Michael,
I concur with your thoughts and it goes back to the fundamental question: what is creativity, what is the process that leads to new ideas and is it something that machines can do? The current conversation around AI is really about LLMs and their interaction as "agents". As such, these are good at absorbing vast amounts of human-produced (and now fed-back machine-produced) writing. However, this comes with the caveat that "new ideas" are applications of patterns that the model has been trained on. Thus, is creativity then mere pattern-recognition or is there more to it? I think that many of us want to believe that there is, but this will not prevent any profit-oriented company from saying otherwise (thank you Ando for rectifying the use of Marx here!).
This leads to a, may I say, spiritual question that is lurking here: is maths a game that can be won by machines that solve problems with no end, and may even ask questions with no end, at a pace that no human can keep up with so that maths should no longer be a human endeavour? There are efforts of many mathematicians (and those that essentially do maths in other fields) to address the issues arising from LLMs being capable of solving maths problems (https://leidendeclaration.ai/, https://mathandai.org/). However, just as (financial) capitalism is a system set on self-preservation, no matter the costs, A"I" will be reinforced by companies pushing for it and researchers forced to by competition for funding and position. I see difficult times for our fields and the people in it ahead, let alone the general societal and environmental repercussions and damage this has and will only increasingly have.
My apologies for this sceptical reply, but hope that we are not blindly endorsing a technological solution to the academic issues that we have and keep technology what it is: a tool to solve a particular problem.
Best,
Henning
PS: I am reminded of mathematics as a mental process that was used by the intuitionist school in order to develop things like Kripke semantics and the philosophy in general. I wonder what the people from that time would have to say.
PPS: For those not on the types mailing list, our esteemed colleague James McKinna made the good point that this discussion cannot be separated from the political questions and suggested some nice reading material:
* Simon Schaffer's speech <https://erasmusprijs.org/en/laureates/simon-schaffer/acceptance-speech/>
* and U. Chicago Press "Working Knowledge: A Simon Schaffer Reader"
* Randy Pollack's "How to believe a machine-checked proof"
On 10/11/2026 21:21, Michael Barr, Prof. via Categories wrote:
I've given some thought to this and I just don't see what AI is going to train on.
For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42^nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly.
For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results.
As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that.
Michael ------------------------------------------------------------------------ *From:* Dusko Pavlovic via Categories <categories-list@categories.org.au
*Sent:* Monday, November 9, 2026 3:58 AM *To:* Michael Barr, Prof. via Categories <categories-list@categories.org.au> *Cc:* Dusko Pavlovic <duskgoo@gmail.com> *Subject:* [categories] Re: Comments on AI and math thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments
I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to
be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on
regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there,
I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to
categories-list-leave@categories.org.au _______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Dear All, As a young mathematician I undoubtedly lack experience influencing the trajectory of this conversation, but I would like to distinguish a related line of questioning: if there is nothing cognitively distinguishing LLM reasoning from human mathematical reasoning, what reasons might we have to prefer mathematics produced from human reasoning? A strong point-of-view, to me, is the question of taste. If we view creative mathematics as an act of complex optimisation under subjective constraints, then it is clear that LLMs have some advantages and some disadvantages: the clearest advantage is the access to greater compute and a greater knowledge space, which creates the *dis*advantage that LLMs have no need for the highly selective judgements people are good at. In a world where artificial intelligence shares more of our reasoning constraints, it is still likely that the aesthetics of LLM-produced mathematics pull in directions which might not be significant to us in the same way. If we are producing work for the edification of other humans, then even a perfect simulation of human reasoning from a simulated human does not accomplish all that we ask mathematics to be. For if mathematics is about human-to-human knowledge and skill transferal, then mathematics divorced, in any way, from the human experience no longer performs the developmental and communal role. In other words, I raise that the question we should be asking is not 'what can AI do', but 'who is mathematics for?' Corey P.s. I encourage all to remember that after asking whether we do/do not adopt a tool, there are still many things within our control: Who provides the tool? What functions do we use the tool for? What restrictions and practices produce the most meaningful alignment? To be active stakeholders in the decision-making process, we need to think about these problems quickly, carefully, and independently (preserve our individual stake as well as those we represent). On Thu, 12 Nov 2026 at 19:01, Henning Basold via Categories < categories-list@categories.org.au> wrote:
Dear Michael,
I concur with your thoughts and it goes back to the fundamental question: what is creativity, what is the process that leads to new ideas and is it something that machines can do? The current conversation around AI is really about LLMs and their interaction as "agents". As such, these are good at absorbing vast amounts of human-produced (and now fed-back machine-produced) writing. However, this comes with the caveat that "new ideas" are applications of patterns that the model has been trained on. Thus, is creativity then mere pattern-recognition or is there more to it? I think that many of us want to believe that there is, but this will not prevent any profit-oriented company from saying otherwise (thank you Ando for rectifying the use of Marx here!).
This leads to a, may I say, spiritual question that is lurking here: is maths a game that can be won by machines that solve problems with no end, and may even ask questions with no end, at a pace that no human can keep up with so that maths should no longer be a human endeavour? There are efforts of many mathematicians (and those that essentially do maths in other fields) to address the issues arising from LLMs being capable of solving maths problems (https://leidendeclaration.ai/, https://mathandai.org/). However, just as (financial) capitalism is a system set on self-preservation, no matter the costs, A"I" will be reinforced by companies pushing for it and researchers forced to by competition for funding and position. I see difficult times for our fields and the people in it ahead, let alone the general societal and environmental repercussions and damage this has and will only increasingly have.
My apologies for this sceptical reply, but hope that we are not blindly endorsing a technological solution to the academic issues that we have and keep technology what it is: a tool to solve a particular problem.
Best,
Henning
PS: I am reminded of mathematics as a mental process that was used by the intuitionist school in order to develop things like Kripke semantics and the philosophy in general. I wonder what the people from that time would have to say.
PPS: For those not on the types mailing list, our esteemed colleague James McKinna made the good point that this discussion cannot be separated from the political questions and suggested some nice reading material:
* Simon Schaffer's speech <https://erasmusprijs.org/en/laureates/simon-schaffer/acceptance-speech/>
* and U. Chicago Press "Working Knowledge: A Simon Schaffer Reader"
* Randy Pollack's "How to believe a machine-checked proof"
On 10/11/2026 21:21, Michael Barr, Prof. via Categories wrote:
I've given some thought to this and I just don't see what AI is going to train on.
For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42^nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly.
For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results.
As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that.
Michael ------------------------------------------------------------------------ *From:* Dusko Pavlovic via Categories <categories-list@categories.org.au
*Sent:* Monday, November 9, 2026 3:58 AM *To:* Michael Barr, Prof. via Categories <categories-list@categories.org.au> *Cc:* Dusko Pavlovic <duskgoo@gmail.com> *Subject:* [categories] Re: Comments on AI and math thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments
I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to
be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on
regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there,
I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to
categories-list-leave@categories.org.au _______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Hi all, I would say there is a lot that is cognitively distinguishing LLM reasoning from human mathematical reasoning. Namely: caring about the question. AI will silently replace your question with another one, offer a miraculous solution that later turns out to be "natural except on XYZ", have no regard for elegance, etc. It follows evals, and has none of its own aesthetics. I love using AI for math—don't get me wrong, it helps me tremendously—but it is very cognitively different. Indeed, I claim that Turing machines don't care <https://topos.institute/blog/2026-09-22-who-cares-about-a-computation/>, and that caring is cognitively important, because it leads to different solutions than non-caring work does. As Robert Pirsig said, *quality* is the external face of what internally is experienced as *care*. This explains why even their proofs have the quality of "slop". Best, David On Fri, Nov 13, 2026 at 1:53 PM Corey Lionis via Categories < categories-list@categories.org.au> wrote:
Dear All,
As a young mathematician I undoubtedly lack experience influencing the trajectory of this conversation, but I would like to distinguish a related line of questioning: if there is nothing cognitively distinguishing LLM reasoning from human mathematical reasoning, what reasons might we have to prefer mathematics produced from human reasoning?
A strong point-of-view, to me, is the question of taste. If we view creative mathematics as an act of complex optimisation under subjective constraints, then it is clear that LLMs have some advantages and some disadvantages: the clearest advantage is the access to greater compute and a greater knowledge space, which creates the *dis*advantage that LLMs have no need for the highly selective judgements people are good at. In a world where artificial intelligence shares more of our reasoning constraints, it is still likely that the aesthetics of LLM-produced mathematics pull in directions which might not be significant to us in the same way.
If we are producing work for the edification of other humans, then even a perfect simulation of human reasoning from a simulated human does not accomplish all that we ask mathematics to be. For if mathematics is about human-to-human knowledge and skill transferal, then mathematics divorced, in any way, from the human experience no longer performs the developmental and communal role.
In other words, I raise that the question we should be asking is not 'what can AI do', but 'who is mathematics for?'
Corey
P.s. I encourage all to remember that after asking whether we do/do not adopt a tool, there are still many things within our control: Who provides the tool? What functions do we use the tool for? What restrictions and practices produce the most meaningful alignment? To be active stakeholders in the decision-making process, we need to think about these problems quickly, carefully, and independently (preserve our individual stake as well as those we represent).
On Thu, 12 Nov 2026 at 19:01, Henning Basold via Categories < categories-list@categories.org.au> wrote:
Dear Michael,
I concur with your thoughts and it goes back to the fundamental question: what is creativity, what is the process that leads to new ideas and is it something that machines can do? The current conversation around AI is really about LLMs and their interaction as "agents". As such, these are good at absorbing vast amounts of human-produced (and now fed-back machine-produced) writing. However, this comes with the caveat that "new ideas" are applications of patterns that the model has been trained on. Thus, is creativity then mere pattern-recognition or is there more to it? I think that many of us want to believe that there is, but this will not prevent any profit-oriented company from saying otherwise (thank you Ando for rectifying the use of Marx here!).
This leads to a, may I say, spiritual question that is lurking here: is maths a game that can be won by machines that solve problems with no end, and may even ask questions with no end, at a pace that no human can keep up with so that maths should no longer be a human endeavour? There are efforts of many mathematicians (and those that essentially do maths in other fields) to address the issues arising from LLMs being capable of solving maths problems (https://leidendeclaration.ai/, https://mathandai.org/). However, just as (financial) capitalism is a system set on self-preservation, no matter the costs, A"I" will be reinforced by companies pushing for it and researchers forced to by competition for funding and position. I see difficult times for our fields and the people in it ahead, let alone the general societal and environmental repercussions and damage this has and will only increasingly have.
My apologies for this sceptical reply, but hope that we are not blindly endorsing a technological solution to the academic issues that we have and keep technology what it is: a tool to solve a particular problem.
Best,
Henning
PS: I am reminded of mathematics as a mental process that was used by the intuitionist school in order to develop things like Kripke semantics and the philosophy in general. I wonder what the people from that time would have to say.
PPS: For those not on the types mailing list, our esteemed colleague James McKinna made the good point that this discussion cannot be separated from the political questions and suggested some nice reading material:
* Simon Schaffer's speech <https://erasmusprijs.org/en/laureates/simon-schaffer/acceptance-speech/>
* and U. Chicago Press "Working Knowledge: A Simon Schaffer Reader"
* Randy Pollack's "How to believe a machine-checked proof"
On 10/11/2026 21:21, Michael Barr, Prof. via Categories wrote:
I've given some thought to this and I just don't see what AI is going to train on.
For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42^nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly.
For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results.
As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that.
Michael ------------------------------------------------------------------------ *From:* Dusko Pavlovic via Categories < categories-list@categories.org.au> *Sent:* Monday, November 9, 2026 3:58 AM *To:* Michael Barr, Prof. via Categories <categories-list@categories.org.au> *Cc:* Dusko Pavlovic <duskgoo@gmail.com> *Subject:* [categories] Re: Comments on AI and math thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments
I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to
be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich
Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on
regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there,
I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated
what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to
categories-list-leave@categories.org.au _______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
If people start talking to AI as well as to other people, using it as a sounding board for their own ideas (which is already somewhat helpful), it will train on those interactions. People will always prefer to talk to other people, since that’s more useful and more fun. But if some experienced people also bounce their ideas off chatbots (because they are physically isolated from their peers, or just out of convenience) then the training corpus will gradually grow. Whether AI will actually learn how to come up with new, meaningful questions on its own is unclear. But the data will be there, so I wouldn’t rule it out. Wesley Sent from my iPad
On Nov 11, 2026, at 12:47 AM, Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
I've given some thought to this and I just don't see what AI is going to train on.
For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly.
For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results.
As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that.
Michael From: Dusko Pavlovic via Categories <categories-list@categories.org.au> Sent: Monday, November 9, 2026 3:58 AM To: Michael Barr, Prof. via Categories <categories-list@categories.org.au> Cc: Dusko Pavlovic <duskgoo@gmail.com> Subject: [categories] Re: Comments on AI and math
thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au _______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Dear Michael and others, On the question of which dataset could AI train on in order to generate questions, I think the answer is quite obvious: that could simply be the prompts that AI users are sending to the model. People, including mathematicians and other scientists, are sending questions to AI on all kinds of subjects everyday. If those questions are used to train a future iteration of the model (and I cannot imagine that this is not what is actually happening), it seems quite clear to me that this future iteration will have a fairly good idea of what constitutes a good and fruitful question. Also, users do not just submit questions to their favorite model: a lot of them use it as a virtual friend and confident and chat about everything and nothing with it. So it is exposed to more human thought processes than any individual human has ever been… Best, Jérôme De : Michael Barr, Prof. via Categories <categories-list@categories.org.au> Envoyé : 10 novembre 2026 15:22 À : Dusko Pavlovic via Categories <categories-list@categories.org.au> Cc : Dusko Pavlovic <duskgoo@gmail.com>; Michael Barr, Prof. <barr.michael@mcgill.ca> Objet : [categories] Re: Comments on AI and math I've given some thought to this and I just don't see what AI is going to train on. For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly. For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results. As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that. Michael ________________________________ From: Dusko Pavlovic via Categories <categories-list@categories.org.au<mailto:categories-list@categories.org.au>> Sent: Monday, November 9, 2026 3:58 AM To: Michael Barr, Prof. via Categories <categories-list@categories.org.au<mailto:categories-list@categories.org.au>> Cc: Dusko Pavlovic <duskgoo@gmail.com<mailto:duskgoo@gmail.com>> Subject: [categories] Re: Comments on AI and math thank you for interesting examples, mike :) while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos. a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols. the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that. if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers... i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities... 2c, -- dusko On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au<mailto:categories-list@categories.org.au>> wrote:
While this is not about categories specifically I have some comments I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au<mailto:categories-list@categories.org.au> To unsubscribe send an email to categories-list-leave@categories.org.au<mailto:categories-list-leave@categories.org.au>
_______________________________________________ Categories mailing list -- categories-list@categories.org.au<mailto:categories-list@categories.org.au> To unsubscribe send an email to categories-list-leave@categories.org.au<mailto:categories-list-leave@categories.org.au>
I find the way you discribe your research very beautiful and inspiring. Yes, that is how I would like mathematics to be. But increasingly, maybe less in math, research teams are letting ai record all research meetings (via zoom or claude for example) to help with summaries and organization. The data thus collected could be used in the way Dusko described.
On Nov 10, 2026, at 12:21 PM, Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
External Message
I've given some thought to this and I just don't see what AI is going to train on.
For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly.
For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results.
As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that.
MichaelFrom: Dusko Pavlovic via Categories <categories-list@categories.org.au> Sent: Monday, November 9, 2026 3:58 AM To: Michael Barr, Prof. via Categories <categories-list@categories.org.au> Cc: Dusko Pavlovic <duskgoo@gmail.com> Subject: [categories] Re: Comments on AI and math thank you for interesting examples, mike :)
while they will probably be echoing in our heads in the coming days, i think it is safe to say at once that AI bots can in principle be post-trained to be as good at asking questions as they are at answering them --- PROVIDED that samples of the process of question generation can be found, something like your and kleisli's conversations. say if they were recorded as emails or videos.
a PRE-trained language model does just one thing: it generates streams of tokens extending given contexts. given a question, a merely PRE-trained model does not try to answer it, but guesses a continuation of the question itself. to recognize a question as a question, and to try to guess an answer, it must be POST-trained on samples of question-answer protocols.
the models solving all those math problems have been POST-trained on humongous corpuses of processes of solving math problems. answering questions. that is why they are good at answering questions. to also be good and questioning answers, they need to be POST-trained on samples of that.
if we want them to help us ask questions, we need to be able to teach them on LOTS of examples of question generation --- like those that you just provided. or maybe not just examples of questioning, but of the *attitude* that generated kleisli's and freyd's questions and ideas: how they "shake up" each construction to find paths beyond it. if we could find enough such examples, we could not only learn from them ourselves, as we do from the examples you just posted, but we could also process them on scale and post-train some helpers...
i am sure there will be lots of people who will now say that we should keep our question-generation processes to ourselves, maybe also all other thinking processes. and maybe they are right. or maybe we should consider models of our thinking processes as a part of ourselves, like we consider our writings as a part of our identities...
2c, -- dusko
On Sun, Nov 8, 2026 at 3:32 PM Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote:
While this is not about categories specifically I have some comments I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
NOTE: This email originated from outside Chapman’s network. Do not click links or open attachments unless you recognize the sender and know content is safe.
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
Dear Mike Face-to-face research is very important. In the 1970s, Max Kelly and I convinced the Australian Research Council and our universities that face-to-face contact to mathematics was as important as a visit to a telescope for an astronomer or a visit to some immovable rock by a geologist. We were able to obtain some funding for our long trips from Australia and for mathematical visitors to Australia. However, in those no-email days we also had a superb postal system. An aerogramme posted in a suburb of Sydney on Monday morning would arrive in my mailbox at U Illinois (Champaign-Urbana) on Wednesday. A lot of my work with Day, Kelly, Freyd and others was done that way. We really thought before we wrote since there was no erasing, not much space, and a week's turn around time. Then there were the long phone calls. These were expensive for international (trunk) calls. One time, talking to Bob Walters for quite awhile, we were unaware that we had a crossed line. A young person listening in recognised that all this Yoneda business was mathematics. They broke in to our conversation saying: "x equals minus b plus or minus the square root of b squared minus four a c all over two a" then slammed their phone down. For me, somehow, the phone worked better than zoom meetings do now; however, zoom is better than nothing. I think the main, and maybe terrifying, things about AI are its speed and arrogance. In high school, we had a little book of tables including base 10 logarithms. No need to know how to multiply anymore! BUT, you still had to have a feeling for numbers to tell when you came up with something ridiculous. In practical statistics classes as an undergrad, we had big heavy machines where you turned the handle 10 times to multiply by 10. They were soon replaced by electricity powered versions. Then there were pocket calculators. Engineers didn't need to do mathematics anymore, people said. Computers. Personal computers. Workstations. And so on. Maybe quantum computers. We can certainly keep doing the old stuff faster. No doubt AI will be used to justify sacking workers of all kinds, not just mathematicians. Will that help make a better world? Time to stop rambling. Ross On 11 Nov 2026, at 7:21 am, Michael Barr, Prof. via Categories <categories-list@categories.org.au> wrote: I've given some thought to this and I just don't see what AI is going to train on. For all my career, I have counted on face to face meeting with people. I wanted to talk to them and the conversations were never recorded. I remember running accidentally into Jon Beck in the 42nd Street Times Square subway station during the winter break 1964-65 and he started telling my about acyclic models. He had just visited Harry Appelgate who had written his PhD on the categorification of that concept. Jon and I eventually modified it further and we were off. We had spent the entire fall semester trying to relate Eilenberg-Mac Lane group cohomology to the free group triple and this solved it instantly. For about 15 years, John Kennison would spend 6 weeks in Montreal every spring and he and Bob Raphael and I would meet for 4 or 5 hours every day and talk, often going to the board to explain, but mainly just talking. Practically every year we wrote a joint paper as a result. This ended with the pandemic and John canceled in 2020. We have not met since. We tried zooming together but that has not worked. We enjoy the conversations but no mathematics results. As far as I am concerned, nothing can replace these face to face sessions and I cannot see training AI on that. Michael
Dear Mike, It's nice to hear from you! I hope you are well. Here, in Switzerland, all is well, as we are relatively protected from _some_ of the world's turmoil - and the scenery remains beautiful! :-) I am writing to say I really appreciated your examples: they put forth convincing, concrete and therefore helpful, examples of what others have said more theoretically. Thank you for enriching the discussion! Best and sincere regards, Gavin On Mon, Nov 9, 2026, at 02:19, Michael Barr, Prof. via Categories wrote:
While this is not about categories specifically I have some comments I would like to make public and this would seem to be my outlet. So please bear with me.
The main thing I want to say about math and AI is that AI appears to be excellent at answering questions, but I see no evidence that it is capable of asking them. And the latter, I feel, is the main thing a research mathematician contributes to the advancement of mathematics. I am going to illustrate this point by two examples from my own career.
I spent a month in Fribourg in 1997 and during that month, Heinrich Kleisli and I wrote a paper called Topological Balls. The way this paper came about was that every day Heinrich and I would meet in his office and discuss the topic and he would ask a question. I would go back to the apartment I was occupying and answer his question. The next day I would explain my answer and we would have a further discussion and he would raise a new question. And I would answer that in the even and so on and so on till at the end of my stay we had a paper. In my opinion, Heinrich contributed more to it than I did. No questions, no paper.
During the year 1969-70, my second at McGill, I was working on regular and exact categories, quite consciously trying to adapt the ideas of abelian categories to non-additive ones. After finding what still seem like the correct definitions, I wanted to adapt theorem that a small abelian category has a faithful functor to Ab. By that time Saul Lubkin's proof and a quite different one from Peter Freyd were known. There was also Barry Mitchell's full embedding into a category of modules over a ring. I succeeded in adapting Lubkin's proof to show that a small regular category had a faithful functor to Set, provided that the terminal object had no proper subobject, except possibly an empty object, meaning one that was not the codomain of any map other its own identity (or any other isomorphism to be sure). In the meantime Kleisli had invited me to spend the year 1970-71 in Fribourg and had gotten a grant from the Swiss Research Council to do so.
So, I took an unpaid leave and headed to Fribourg, On the way there, I stopped in Philly to visit my family and looked in on Peter Freyd at Penn. I explained what I had done and Peter asked me whether it was possible to get a full embedding theorem out of the method by replacing Set by M-Set for a suitable monoid M (which turned out to be the monoid of endomorphisms of the original embedding). To my surprise, it was! And not particularly hard to prove. That means that Lubkin could have had a full embedding theorem too, had he realized. Anyway, on to Fribourg where I showed Heinrich what I had done and he asked whether it would be possible to drop the condition on subobjects of 1 and replace M-Set by a category of the form Set^C for some small category C. It was! And not too hard either. The point was that, unlike me, he thought of a monoid as a small category.
So two questions, one from Peter and one from Heinrich, had elevated what was probably a minor result to one of the major accomplishments of my career. And this is why I think that questions are more important than answers.
Michael
_______________________________________________ Categories mailing list -- categories-list@categories.org.au To unsubscribe send an email to categories-list-leave@categories.org.au
participants (15)
-
Ando Razafindrakoto -
Corey Lionis -
David Spivak -
Dusko Pavlovic -
gavin_seal@fastmail.fm -
Henning Basold -
Johnathon Taylor -
Jérôme Fortier, Dr -
Kurz, Alexander -
Michael Barr, Prof. -
posina venkata rayudu -
Posina Venkata Rayudu -
Ross Street -
Vyacheslav Kalmykov -
Wesley Phoa