Because it succinctly but correctly spotlights the motivation for the release and the fact that motivation is self-serving and not necessarily a positive for the field.
You’re talking about whether something can be verified. I’m talking about the mental DDOS releasing work like this creates
The navier stokes paper is 166 pages long and the lean code is over 500 kloc. Nothing about that is “clear”. Someone has to real ALL of that to see if the proof is clearly right or wrong.
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 — 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.
Looks like mathematicians are going through the same stages of grief that software engineers went through in the last 1-2 years.
Tao is obviously much smarter than I'm and he knows a lot more about how math research works than I do. However, from a layman's view from 10,000 feet above, these statements look totally nuts. They're basically telling OpenAI or any AI lab to stop using the models for advanced math research so the humans can continue to do it slowly instead. This feels similar to Frank Herbert's Dune novels on banning computers.
What they want won't work for more than a few months at longest. There will be many mathematicians who disagree with this, will use OpenAI/LLMs to advance cutting edge math research, and the mathematicians who don't use it will be left behind.
> [This is a guest post by the Association for Human Mathematics, reposted from their statements page. This blog post was initially written in a different file format and converted using AI. — T.]
We do have rules in place for animal testing in the pharma industry for more or less ethical reasons, so it's not unheard of use it as an argument to either stop or incentivize a company to do something. I agree that it is nuts from the point of view of actually changing anything, at least now.
It's just an identity and job security crisis for these mathematicians. Very similar to software engineers over the last 1-2 years.
They'll learn to accept reality sooner than later. Software engineers are still employed. In fact, I'm busier than ever since coding agents hit mainstream. My job is just different now but it's for the good overall.
These mathematicians will also figure this out.
Asking OpenAI to stop testing advanced math on internal models and stop discovering solutions is just crazy talk.
Yes and if they were keeping mathematicians in cages that would be a good point. These are the same mathematicians who already agreed to work with OpenAI, who were apparently okay with how the models were trained but were concerned about how they night be used. Which sounds fair in the abstract, but concretely it means that they didn't care about intellectually property rights until it was their own jobs on the line. Many such cases.
Did you even read the quoted text you posted on here?
They say:
> 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
Read: "on internal models"
Then you go:
> They're basically telling OpenAI or any AI lab to stop using the models for advanced math research because for some reason it isn't "ethical".
18 comments
[ 1.6 ms ] story [ 13.6 ms ] threadOne of the best sentences in this short and potent article.
The navier stokes paper is 166 pages long and the lean code is over 500 kloc. Nothing about that is “clear”. Someone has to real ALL of that to see if the proof is clearly right or wrong.
Tao is obviously much smarter than I'm and he knows a lot more about how math research works than I do. However, from a layman's view from 10,000 feet above, these statements look totally nuts. They're basically telling OpenAI or any AI lab to stop using the models for advanced math research so the humans can continue to do it slowly instead. This feels similar to Frank Herbert's Dune novels on banning computers.
What they want won't work for more than a few months at longest. There will be many mathematicians who disagree with this, will use OpenAI/LLMs to advance cutting edge math research, and the mathematicians who don't use it will be left behind.
> [This is a guest post by the Association for Human Mathematics, reposted from their statements page. This blog post was initially written in a different file format and converted using AI. — T.]
https://www.ahmath.org/statements
They'll learn to accept reality sooner than later. Software engineers are still employed. In fact, I'm busier than ever since coding agents hit mainstream. My job is just different now but it's for the good overall.
These mathematicians will also figure this out.
Asking OpenAI to stop testing advanced math on internal models and stop discovering solutions is just crazy talk.
They say:
> 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
Read: "on internal models"
Then you go:
> They're basically telling OpenAI or any AI lab to stop using the models for advanced math research because for some reason it isn't "ethical".
Maybe it's their use of internal models?