Who is the target for OpenAI’s “publications”?

I believe that not enough mathematicians have understood the scale and purpose of the actions of LLM companies. Let me be perfectly clear: their goal is not to advance science, and it is especially not to empower scientists. Their goal is to divert research and education budgets away from scientists, and in their pocket.

The marketing stunts they perform by feeding the arXiv to their statistical parrots are not aimed at “empowering scientists”, let us stop pretending as much. To empower scientists, they would have to give us better tools, or at least provide us value, and their slop-posting does not hold any value.1 Mathematicians are not stupid, we understand that, so of course most of us are incredibly enraged, which is in my opinion the only appropriate reaction. What we don’t seem to understand enough, though, is that these are not good-faith efforts to impress us. In fact, this is not an effort to impress us at all – we were never the target of this marketing campaign. The targets are the uninformed, unscientific policymakers who fund us!

These companies were never after our IT budgets, they want our entire funding, and to get it they are forcing their way into newspaper covers with sensationalist titles they can afford to pay for. The message they want to send is that, for a university, “it would be way better to spend 20k€ a month on tokens, instead of employing those dipshit mathematicians, who spend their time drinking coffee and snorting chalk”.2 After all, look at how many theorems these tokens are able to prove!. Stop paying them and pay us. Stop funding and forming talented members of your society, pay us forever to run GPUs instead! We can produce a4 documents that look like mathematics too!

They buy credibility from us

This message will not resonate with us, of course – we understand that it is stupid – but their goal is to have it resonate with policymakers. To ensure it does, these companies keep hiring prestigious mathematicians, leveraging their scientific standing and credibility for marketing. I do not understand why a tenured mathematician would play into this game: are we all this naïve?

I am sorry, but this needs to be said: do we really think that OpenAI pays Tsimerman for his deep insight in transformer architectures? Do we think that they collaborated with Tao because he is a very good machine learning engineer? Maybe they keep giving money and tokens to Alpöge because he is fun at parties? Let us be serious.

If you are a mathematician, and an AI company wants you to work with them, I have sad news for you: they don’t care about your talent, your expertise or your opinions. They don’t care about your math, most importantly. Maybe they put you to work on something you find interesting, maybe they just told you “here is unlimited tokens, use them as you please and cite us”, I don’t know. Maybe you don’t either. Maybe you are part of some advisory committee that they begged you to create. Maybe you spent days writing guidelines on how they should publish their slop, as if they ever gave a shit. As if they don’t already have marketing departments whose entire purpose is to coordinate publications in order to maximise attention grabbing.

They are not doing this because your work is useful to them. They are doing this because your reputation is!3 All they care about is to be able to say “look, these famous mathematicians work with us! It is the others who are wrong, fire them and give us the money”. You are helping them divert the money that pays you, no matter what you think they onboarded you for. If and when that happens, they will dump you too. They will have replaced your department with chatbots by then.

In conclusion

AI companies are pursuing a very specific goal: diverting public research budget to themselves. It is not a coincidence that we are seeing, at the same time, a push for replacing teaching and TA duties with chatbots: these are joint campaigns to simultaneously devalue what actual researchers offer to universities, as I wrote already, and inflate the “value” of the product they want to trade against our salaries.

We must, as a community, understand this fact and act accordingly. Ignoring the people who want you dead is not an option, if you want to live.

  1. If you are a mathematician and you actually think that the 700+ writeups OpenAI put out recently hold any value, I would very much like to hear your opinion on why they do. My position is that they do not, because they are written like shit, they copy from the literature without citing their sources, and most importantly because no one cared enough to write them, so I don’t see why anyone should care to read them. I certainly won’t. ↩

  2. On a side note, I am told that this is what we look like to non-scientists. Maybe part of the efforts we make to defend our profession from AI salesmen should be spent on improving our public image, by the way. Might make things easier for us. ↩

  3. If you don’t think so, please see why that might be. ↩