Proofs and Prompts

AHM Statement on OpenAI’s October 6 Release of Mathematical Documents

AHM communications group

Yesterday, on October 6th, 2026, OpenAI – which is currently defending lawsuits against accusations of illegal plagiarism, copyright infringement, and trademark dilution – released a repository of manuscripts purporting to contain solutions to a number of high-profile problems in mathematics.

Mathematicians did not ask for this work to be done. The Advisory Group on Mathematics and Artificial Intelligence, from whom OpenAI has claimed to derive its legitimacy, opened their initial advisory statement by saying that frontier AI corporations should not test advanced mathematical problems on internal models. In ignoring the central premise of the Advisory Group’s position, OpenAI has indicated total disregard for the norms of scientific research —\text{\textemdash} norms that guarantee that mathematics remains trustworthy, ethically researched, and in the public interest.

Mathematicians have a particular vision of progress that is informed by history and field-specific considerations. We reject OpenAI’s assertion that this release advances our subject, and we urge mathematicians and the public to view the value of this publication model with due skepticism. 

Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power. We urge mathematicians to discontinue their work with OpenAI and to return to a vision of science that centers human understanding.

Crossposted from the AHM website.

Received 7 October 2026.

42 responses to “AHM Statement on OpenAI’s October 6 Release of Mathematical Documents”

  1. Artin Wedderburn Avatar
    Artin Wedderburn

    Hi AHM,

    I was reading your text and got struck by your assertion that “Mathematicians did not ask for this work to be done.”

    I am a professional mathematician and I am partly* glad that OpenAI developed a technology which can solve these problems and that they shared the proofs publicly. Neither you nor the Advisory group represent me.

    I believe that ultimately your reactions have, as your false assertions, nothing to do with human-understanding (whatever this brand new concept means) but everything to do with human emotions. I totally understand the anxiety caused by these developments for my younger colleagues concerning their future prospects*, but for tenured mathematicians going on with the human-understanding flurry (and unless they are only being emphatetic), I’d like to say, get a life.

    Reply
    1. A professional mathematician Avatar
      A professional mathematician

      … said an anonymous OpenAI employee.

      Reply
    2. A tenured mathematician too busy getting a life to write a longer response Avatar
      A tenured mathematician too busy getting a life to write a longer response

      Your younger colleagues, who are anxious about their future prospects because of these developments, will surely be happy to know you’re doing totally fine and that you are in fact partly glad about this state of affairs. Have you considered talking to them every once in a while?

      Reply
    3. lohn_jennon Avatar
      lohn_jennon

      Right, so if I understand correctly, you are a professional mathematician who does not understand what ‘human understanding’ means? Sure ok

      Reply
    4. Niki Avatar
      Niki

      Based on the noble tone of this AHM statement, we can be assured that the ~500 registered sacred guardians of AI-free mathematicians will police humanity’s mathematical progress. How reassuring!

      Reply
    5. Arihant Jain Avatar
      Arihant Jain

      You might be familiar with William Thurston’s paper, On Proofs and Progress in Mathematics. The concept of human-understanding isn’t new. On the other hand, Its current popularity highlights that we may not have fully grasped what it actually means to ‘do’ mathematics in the first place.

      Reply
    6. Pierre Menard, Author of the Quixote Avatar
      Pierre Menard, Author of the Quixote

      “get a life” in this context is equivalent to saying “don’t be sad” to a clinically depressed individual. What an utterly disgusting comment.

      Reply
      1. Anonymous Avatar
        Anonymous

        I agree. I am actually surprised that this comment was allowed to be posted.
        Also, it is a bit ridiculous to say “get a life” if you are dead since 1948 (Joseph Wedderburn) or 1962 (Emil Artin).

        Reply
  2. Jake Avatar
    Jake

    Gotta love a strong opening sentence!

    Reply
  3. Craig Avatar
    Craig

    As a Star Trek fan, I say “Regards to Professor Dunsel”.

    Reply
  4. Sirawit Pongnakintr Avatar
    Sirawit Pongnakintr

    First, as far as I’ve remembered (correct me if I’m wrong), OpenAI has already generated these results **before** reaching out to AGMAI.

    AGMAI says that “If AI labs produce significant mathematical results, they should responsibly release the results, as outlined in this document, as soon as possible” while also says that “[We] prefer the AI labs not to test problems on internal models”.

    My interpretation (if I were someone at OpenAI) is therefore: Yes, we should release these ~400 results immediately, and stop trying to achieve more results later. In this sense I don’t see how OpenAI rejects the AGMAI’s recommendations.

    Now, onto the stronger point, and I’m resisting my urge to say anything rude or offensive: “Mathematicians did not ask for this work to be done.” Who are the “Mathematicians”? Do I have to ask for your permission when I study a topic I like? And if not, why would AI companies have to ask for some *particular* people’s permission? Mathematics is too big to be universally represented by a single group, whether Fields medalists, or AGMAI, or AHM. One thousand people cannot claim the ownership of mathematics. Neither can one million people. Neither can one billion people.

    Everyone has the rights to do mathematics. AI companies have their rights to do mathematics.

    We must know. We will know. I support human understanding as much as I support AI companies to develop new technologies to advance our understanding towards mathematics. Preventing some entity (whether human or AI) from walking further towards the unknown does not seem to be supportive of human understanding, at least, in my opinion.

    Reply
    1. x Avatar
      x

      Let me say this kindly, your arguments are philosophical and do not really look like that of an experienced professional mathematician.

      ‘AI companies have their rights to do mathematics.’

      One can make a very strong argument that AI companies are in fact NOT doing mathematics.

      Reply
    2. lohn_jennon Avatar
      lohn_jennon

      “I support AI companies to develop new technologies to advance our understanding towards mathematics. ”

      Name one problem whose solution you have understood from the 722 manuscripts generated by OpenAI.

      Reply
      1. Sirawit Pongnakintr Avatar
        Sirawit Pongnakintr

        I haven’t read the “722 manuscripts”, so I cannot claim anything personally on that.

        But on other things, yes, I can claim a lot. For example, a few days ago I was reading the Wikipedia on Grothendieck topologies. I was entirely confused when the definition of a sheaf is a presheaf F whose the natural map “Hom(Hom(-, X), F) to Hom(S, F)” is bijective for all X and S. Obviously I don’t want to *destroy* my understanding process, so I tried to work with this on my own for like an hour, and still couldn’t understand how this corresponds to a sheaf in the usual sense, so I asked ChatGPT with this exact prompt:

        > I’m kinda confused about the definition of sheaves on a site. I thought morally it should be a generalization of sheaves on a topological space into the case of Grothendieck topologies. The definition of a presheaf makes sense: for a site $(\mathcal{C}, J)$, a presheaf is just a functor $\mathcal{C}^{op} \to \mathrm{Sets}$. However, in Wikipedia, for a presheaf $F$ to be a sheaf, the natural map $\mathrm{Hom}(\mathrm{Hom}(-, X), F) \to \mathrm{Hom}(S, F)$ for all $X \in \operatorname{Obj}(\mathcal{C})$ and for all $S \in J(X)$. This statement isn’t very easy for me to parse, because right now I see $\mathrm{Hom}(-, X)$ as a presheaf, and I see $\mathrm{Hom}(\mathrm{Hom}(-, X), F)$ and $\mathrm{Hom}(S, F)$ as “the presheaf of presheaf morphisms”, but that the restrictions are inverted, i.e., for $i \colon Y \to Z$ in $\mathcal{C}$, the natural restriction is $\mathrm{Hom}(\mathrm{Hom}(Y, X), F(Y)) \to \mathrm{Hom}(\mathrm{Hom}(Z, X), F(Z))$ given by $f \mapsto f’$ the unique object such that $f’|_Y = (g \mapsto f(g|_Y))$. Is this the correct thing that this map means? Or am I confused by the nested homs? What is the correct definition of sheaves?

        And then ChatGPT helped clarified that the outer “Hom” does not mean a sheaf morphism or anything like that, but it means natural transformations, so what we’re really considering is Nat(h_X, F) to Nat(S, F). I worked a bit more on my own until I reached an “aha” moment where I could see that this is precisely what it means for a presheaf to be “gluable”.

        Back to our discussion, creating hundreds of manuscripts, and training AI on mathematics, to me, are very essential parts of the process towards creating an AI that can teach me mathematics (more correctly, more simpler explanation, more direct diagnostic on my confusion, etc). Therefore, I benefit from this, and it’s a fair business since I pay OpenAI for $100 a month for this service. I would be unhappy if the “Mathematicians” want to stop the progress of some AI entity that could teach me the mathematics I wanted to know. (Especially in this age where professional mathematical supervision is scarce, admissions are extremely competitive, and no one can afford to spend 10 hours with me per day, except AI)

        Reply
        1. Lautaro Avatar
          Lautaro

          Unlike people like you, people like me who live in developing countries and have long ago forsaken the dream of being a mathematician, cannot afford to pay 100 dollars a month to get an education of the sort you describe. As an electrician I make much less.

          Yet even though I could not afford college, I do study mathematics autonomously and do spend a brutal amount of time pondering things without any help other than that afforded by the book I downloaded illegally.

          If I cannot see how to solve some particular problem in Lang (say), I will be stuck with it until it clicks. Even if I could, I wouldn’t ask ChatGPT for clarification or whatnot, since that defeats the whole point.

          So it is a question of ethical fitness (more or less). Mathematics is a human endeavour, and you quoting Hilbert to further apologetic narratives on behalf of these egregiously ill-willed companies is sort of really sad.

          Reply
          1. Sirawit Pongnakintr Avatar
            Sirawit Pongnakintr

            Perhaps you have a prior belief that these are “egregiously ill-willed companies” and I’ll respect your opinions. I, on the other hand, do not have such belief.

            My opinion is that “if $100/month is too expensive, then OpenAI should sponsor an education pack for students who want to use AI to propel their learning trajectories” and not “oppose OpenAI” or “ask OpenAI to stop making progress on their models around mathematical tasks”.

            Mathematics is a human endeavor, yes. I am literally a human trying to study mathematics.

            Humans don’t have wings. We wanted to fly. We can’t. We didn’t give up. We built airplanes. We built spaceships. And we went to the moon.

            I am stupid. I don’t have enough “genius brain power” to do mathematics. I do not give up. AI is my wings. And I’m betting my life on the chance that it may take me to the moon. And if not, may I die in the spaceship, as it’s better than sitting at home crying about the fact that I don’t have wings and cannot go to the moon. (At least this is my take, and I’m quite prepared if you or anyone else disagree, but ok)

            Reply
            1. Lautaro Avatar
              Lautaro

              What I find remarkable in you is that I sort of know where you are coming from and feel sympathy for some of your views. Perhaps in some other circumstance we could even have been friends.
              Anyway, you correctly identify a core belief that we positively DO not share.

              But do not even try to argue that I am sort of “crying” because I do not have wings. That is condenscending and insulting.

              Grothendieck did not need these tools, hell, Gauss, Euler, Newton, Archimedes, etc did not.

            2. Sirawit Pongnakintr Avatar
              Sirawit Pongnakintr

              Alright. You are not crying. If my words are interpreted in a condescending or insulting ways, I apologize in advance.

              I am literally crying right now. I wish I could be smart as you, or Grothendieck, Gauss, Euler, Newton, Archimedes.

              I did not choose to be born stupid. But it is what it is. The path that I choose is to accept such stupidity and try to do everything to go forward. To me, AI is that tool. I agree to disagree, if you believe this is where our arguments should end. Thanks for the conversation.

        2. lohn_jennon Avatar
          lohn_jennon

          lol ok sure. You have not defended why you thought openAI progressed human understanding with these 722 problems, which is precisely what the AHM statement is saying.

          What you’ve claimed is that it taught you about the Yoneda lemma. Good for you but how does that effect human understanding on the large scale?

          Reply
          1. Brandon Yates Avatar
            Brandon Yates

            In one day? Surely that is a normal amount of time to review a paper in your subfield…

            Reply
            1. lohn_jennon Avatar
              lohn_jennon

              The point is the following: if OAI understood at least one of the solutions then sure I would agree that they advanced the field.
              When humans produce a paper, they understand whatever is in the paper. For example you can just question the author. This is true even if they produce it with an LLM.

              On the other hand this huge dump of purported problem solutions which **literally** no one understands is harmful to the community.

              Since the OP of this comment claimed that LLMs were advancing human understanding, I asked them if they understood any paper in the proof dump. If they did, then sure that would advance collective human understanding as *at least* someone understood new math which no one understood before.

              The Yoneda lemma is well understood. While I congratulate the OP on understanding the lemma via ChatGPT, it does not advance human understanding as a whole.

    3. IS Avatar
      IS

      The question isn’t about who has a right to do mathematics. Almost any action can be justified by asserting a right to X.

      The question is whether it is good for OpenAI and others to enact an enclosure of the entirety of human knowledge when training the AI, and then attempt to displace, deskill, or otherwise disempower human workers, including human workers who contributed to the training data without consent or compensation.

      I would also like to dispute that the only goal is “understanding” of mathematics. Mathematics is a job. Generally speaking, I would prefer that the average worker in society retains more labor power even if automation would be “more efficient”. I don’t think the average member of the public wants to have no useful labor to sell, since their income depends on it.

      Reply
      1. Sirawit Pongnakintr Avatar
        Sirawit Pongnakintr

        I can see your point. But, as usual, I think I have a very different fundamental assumptions that it would not be productive to argue to your comment directly. Perhaps I’ll list some of my beliefs here:
        (1) Mathematics is NOT a job. It is an identity. Mathematics is too important for me to be a job. I’m also surprised by how the public thinks about my position. It’s the same thing as “I identify myself as a gay male” or “I identify myself as a transwoman”. It’s more about what one choose to be than how the person is employed by the society.
        (2) I disagree to “have useful labor to sell”. People don’t play chess because the society needs the best chess player to contribute to human lives, and artists don’t draw paintings because they can be sold. You’re claiming this on the “average member of the public”, so I won’t say that you’re wrong or something like that, but I’ll probably say that under this assumption, I’m a minority, and not an “average member of the public”.
        (3) All jobs should be killed (if you think mathematics is a job, then that part of mathematics should also be killed, in my opinion). People should do whatever true to their dreams, not whatever the public wants out of them. AI is a road to that liberation.

        The question you raised is invalid under my assumptions, because I already think “mathematics as a job” should be abolished (as fast as possible). It may still be valid under the “average member of the public”, but at least my comment and opinion stands.

        Reply
        1. IS Avatar
          IS

          “(1) Mathematics is NOT a job. It is an identity. Mathematics is too important for me to be a job. I’m also surprised by how the public thinks about my position. It’s the same thing as “I identify myself as a gay male” or “I identify myself as a transwoman”. It’s more about what one choose to be than how the person is employed by the society.”

          How does the rent get paid if math is an identity? Right now people are able to do math full time because it is a paid position (i.e. a job).

          (2) I disagree to “have useful labor to sell”. People don’t play chess because the society needs the best chess player to contribute to human lives, and artists don’t draw paintings because they can be sold. You’re claiming this on the “average member of the public”, so I won’t say that you’re wrong or something like that, but I’ll probably say that under this assumption, I’m a minority, and not an “average member of the public”.

          Very few people make a living playing chess or creating paintings. I’m guessing you would rather do mathematics full time based on your desire to do a phd in math. But if one cannot make a living doing math, I’d imagine the alternative is to have a day job and do math for a couple hours at night in one’s free time. If ChatGPT for 100$ is so great for learning math and producing research, why would someone pay a human to teach or do research for wildly more money? Your previous reply about learning Grothendieck topology related mathematics already suggests that it is more useful than human written resources, its always available, its less money, etc.

          I understand that lots of mathematicians think we get paid to “advance understanding” or whatever. This is a nice thought. But math gets way more funding than the humanities and certainly the arts and music. Those fields absolutely contribute to society. The difference between math and these fields is that math is useful for corporations to make commercial products, weapons research, university prestige.

          (3) All jobs should be killed (if you think mathematics is a job, then that part of mathematics should also be killed, in my opinion). People should do whatever true to their dreams, not whatever the public wants out of them. AI is a road to that liberation.

          I mean that would be great. But if AI kills jobs the average person has no economic power and therefore no political power. Unions, mass protest, and collective action requires the average person to be able to impose a cost on those in power by withholding labor or not complying. Furthermore, the state historically had constraints because it required average people to enforce the law. AI erodes both of these mechanisms by stripping people of their labor power and potentially automating state actions, including law enforcement. So I think you just get whatever the tiny ruling minority wants, which probably isn’t liberation for the average person based on how things work these days.

          Reply
          1. Sirawit Pongnakintr Avatar
            Sirawit Pongnakintr

            > How does the rent get paid if math is an identity? Right now people are able to do math full time because it is a paid position (i.e. a job).

            Rent isn’t paid, by default, and that’s why I ask for institutions to pay the mathematicians, not because they’re doing their jobs, but because they’re a part of the human flourishment.

            > But if one cannot make a living doing math, I’d imagine the alternative is to have a day job and do math for a couple hours at night in one’s free time. If ChatGPT for 100$ is so great for learning math and producing research, why would someone pay a human to teach or do research for wildly more money?

            ChatGPT can explain me about Grothendieck topologies, but it is not (yet) strong enough to completely direct my research and learning trajectory towards understanding more math (or at least I’m not confident enough to hand over such risky task to ChatGPT today). I prefer to have a human advisor, perhaps to talk to for 1 hour per week, and to have ChatGPT to talk to for 30 hours per week, and to work on my own for like 10 hours a week.

            > But math gets way more funding than the humanities and certainly the arts and music.

            I disagree with the funding policy. Math should be funded as much as humanities, arts, and music.

            > So I think you just get whatever the tiny ruling minority wants, which probably isn’t liberation for the average person based on how things work these days.

            I partly agree with your point, but my opinion is that there will be an economic tension between the AI-powered class and the ordinary people, and either there will be another socialist revolution, or AI companies will gracefully pay more “taxes” (or more politely: “donations”/”philanthropy”) to the people they replaced.

            Reply
  5. quicklyf0e5a9188c Avatar
    quicklyf0e5a9188c

    Here is a company that competes with its customers. I spent 2 months asking one of these conjectures to ChatGPT pro, without an epsilon of progress. Why should I keep paying now?

    Reply
    1. Random Avatar
      Random

      So that you will keep paying for the next more powerful model. Of course they are not going to give you all the good things at once.

      Reply
    2. Angry postdoc Avatar
      Angry postdoc

      I voted for the leopards eating people’s faces party and I NEVER THOUGHT THAT THE LEOPARDS WOULD EAT MY FACE

      Reply
  6. Manuel Avatar
    Manuel

    Obviously there are many different opinions and emotions. Many conflicting opinions show how diverse the community and motivations to pursue a career in science and math research are.

    I think the best is to go with the flow:

    – Its great not everyone uses AI, not all problems will be solved that way. Many important discoveries and future technologies may not be indicated by existing data and everyone using a lot of Ai does not sound like a good way.
    – Its great there are people who have ideas how to responsibly employ Ai to push some of the limits and to expand our knowledge with the help of Ai. Many important complex problems may never be tackled without Ai.
    – Its great that there are people in between these points.

    Clearly, there are serious potential issues about plagiarism and theft in association with Ai use. There are also many other serious issues, for instance training of students, as reliance on Ai can lead to a lack of actual learning, there seems no way to speed up learning and understanding of individuals beyond what existing first class programs and lecture materials could already provide, we have a natural and healthy bottleneck, or lets call it a human neck. We need to engage over long period with something using various mechanisms of our own brain in order to learn something.

    So there is a challenge. A big one. But mathematicians like challenges. We can find ways to navigate this. We can find ways to evaluate progress. We can solve harder problems with Ai that Ai could not solve without us, there are always harder and more relevant problems. We can also have entirely novel and groundbreaking ideas without Ai, and this will happen. We may need to think more about how to share our data and knowledge with automated systems, we may need to think about what are responsible contributions that have scientific value and how to evaluate them. But it can be done, I have no doubt. Just its a challenge, something to struggle with, to fail with, to try again, and so on.

    Reply
    1. Anonymous Avatar
      Anonymous

      How can one responsibly use genAI? Data centers are destroying the environment and the climate. Big Tech is destroying democracy. Although a limited use of SLM or even more basic forms of AI makes sense in some very specific fields (such as medicine), the spread of genAI and the increasing power wielded by Big Tech companies is a dangerous threat.

      Reply
  7. Nim Chimpsky Avatar
    Nim Chimpsky

    “[…] OpenAI has indicated total disregard for the norms of scientific research – norms that guarantee that mathematics remains trustworthy, ethically researched, and in the public interest.”

    Please explain what you mean by “trustworthy”. If you question the correctness of these new proofs, please indicate how in light of the Lean certificates. And more to the point: How is the norm that human understanding lies at the heart of mathematics “in the public interest”? Yes, nobody becomes a mathematician because they want only a yes/no answer to the Riemann Hypothesis, but mathematics is funded by the public because the public expects applications in the real world somewhere downstream of pure math research – not as occupational therapy for mathematicians. If you wish to not destroy mathematics (i.e., degrade it to a hobby or an occuption funded by mathematically inclined philantropic billionaires), it might be worth considering who pays the bills. A productive way forward would be to push publicly-owned AI in order to decouple the new center of mathematical (and other) power from corporate interests. You might also consider channelling your anger, fear and disappointment towards advocating for UBI, instead of towards demonization of the technology that has (unfortunately for us) destroyed the reality to which you cling so much. That would truly be in the public interest and allow everyone to continue doing math in whatever way they please.

    Reply
    1. Pierre Menard, Author of the Quixote Avatar
      Pierre Menard, Author of the Quixote

      > applications in the real world

      FFT complexity from O(n log(n)) to O(n log(n)^{1 – 1e-13}) is truly going to change the world and cure cancer.

      Reply
      1. Nim Chimpsky Avatar
        Nim Chimpsky

        Do you really believe that the mathematics you do is utterly pointless, apart from the joy it gives you and the few dozen other experts in your particular field? If so, I truly feel sorry for you.

        I would invite you to state so in future grant applications. Except that this wouldn’t just sabotage your own career, but also the careers of those of us who aren’t so unwilling to face up to the new reality. You seem to have an some literary interest, have you read Ibsen’s the wild duck? It might apply here.

        Reply
    2. Lemon Avatar
      Lemon

      >A productive way forward would be to push publicly-owned AI in order to decouple the new center of mathematical (and other) power from corporate interests.

      Will never happen and you know it.
      Mathematicians have too little power to do something with AI other that strongly worded letters.

      Reply
      1. Hume Chang Avatar
        Hume Chang

        If Jim Simons was still alive and agreed with them, he could donate a data center to them to run some Chinese open source models.

        Reply
    3. Anonymous Avatar
      Anonymous

      “If you question the correctness of these new proofs, please indicate how in light of the Lean certificates” May I suggest you have a look at this paper? https://arxiv.org/abs/2610.08144

      Reply
      1. Nim Chimpsky Avatar
        Nim Chimpsky

        Anybody who joined the AHM because they fear that these new proofs are the result of Lean hacking – which I assume is nobody – is exempt from my criticism.

        Reply
    4. Lemon Avatar
      Lemon

      >How is the norm that human understanding lies at the heart of mathematics “in the public interest”?

      Because only understanding guarantees correct application?

      Reply
      1. Nim Chimpsky Avatar
        Nim Chimpsky

        Imagine OpenAI announced tomorrow that their internal model found a cure for cancer, and that the association for human pharmacology responded by demanding OpenAI withhold this result, because “only understanding guarantees correct application”, and “understanding” presumably requires waiting until some human researcher developed the cancer curing drug without AI assistance. How would the public respond?

        Reply
    5. lohn_jennon Avatar
      lohn_jennon

      They have literally withdrawn 3 papers (at current count). What are you really talking about??
      This is what is meant by trustworthy.

      Reply
    6. lohn_jennon Avatar
      lohn_jennon

      Oh my god ‘ How is the norm that human understanding lies at the heart of mathematics “in the public interest”?’

      This statement is so misguided that I don’t know what to say. What exactly do you think advances mathematics? Do you really think a true/false answer to the Riemann Hypothesis is worthwhile? Please explain exactly how.

      Also this OAI/LLM shilling about highly educated people ‘demanding UBI’ is going to backfire. If you are under the impression that mathematicians (and other highly educated white collar workers) are going to roll over, accept their demise and hand over our agency to be fed by Sam Altman/Elon Musk/ Dario or what have you then you have seriously misunderstood how dangerous discontent can and will be.

      Let me be very clear: we are NOT sheep. If you come for us expect resistance and do NOT cry about it.

      Reply
  8. Bruno Dular Avatar
    Bruno Dular

    Why isn’t anybody *here* discussing the way we, as mathematicians, can contribute to AI safety research?

    Along the previous months and years, it has become increasingly clear that the main, urgent and serious, problem facing society is not whether these math solutions contain plagiarism, not whether such (or any) math outputs have value, not our jobs, and not even the fate of the mathematical community.

    The main, urgent and serious problem is to get a better understanding of these AI models that are developing too fast, too fast even for Anthropic, OpenAI et al., in order to develop robust safety measures and to be able to even have any hope of an optimistic future.

    Nobody is talking about it here, but many people (albeit too few) are already working on it. As mathematicians, we can start contributing and have an impact. I recommend, for example, the following readings/videos:
    – Math for AI safety, by Lionel Levine: https://arxiv.org/pdf/2609.15289
    – MAISI: https://maisi.org/research and https://mathforaisafety.org/resources
    – Hugging Face incident: https://openai.com/index/hugging-face-incident-and-the-road-ahead/
    – Kurzgesagt : AI just crossed the terrifying line, now what? : https://youtu.be/ujkD4SxPKOI?si=N6GXD8sOcuLAMffW
    – Possible future scenarios (not to dismiss as Sci-Fi) : https://ai-2040.com/

    Reply

Comments are moderated. Read our comment policy.

Add to the discussion

Subscribe

New posts by email.

Prefer a feed reader? Subscribe by RSS.

Also on Mathstodon.

Join the discussion

Latest comments across the site.

Timeline of events

Discover more from Proofs and Prompts

Subscribe now to keep reading and get access to the full archive.

Continue reading