Mathematicians Condemn OpenAI's Mass Release of 700+ Math Documents as a Show of Power
AHM Statement on OpenAI's October 6 Release of Mathematical Documents
The Association for Human Mathematics (AHM) has issued a statement condemning OpenAI's October 6 release of over 700 mathematical manuscripts, which purport to solve high-profile problems. AHM points out that OpenAI ignored the Advisory Group on Mathematics and Artificial Intelligence's warning against testing advanced problems on internal models, showing disregard for scientific norms. The statement urges mathematicians to stop working with OpenAI and to prioritize human understanding over what it calls a demonstration of power, not scholarship.
Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power.
- skwirl
I'm having a really hard time seeing how this letter could conceivably be perceived as persuasive. In fact, this letter makes it clear that the people behind it are seeking to maintain their previously entrenched power, prestige, and control over a field of human knowledge. And they are going about it in a futile and debasing way.
- spongebobism
There are 2 strong points buried between the snark: that the AI math advisory group, which OpenAI announced working with just 2 1/2 weeks ago, is against using internal models to solve open problems (per their general guidelines[0]); and that this mode of dumping results before having a friendly formulation doesn't center human understanding.
IMO OpenAI's way of publishing also goes against other parts of the general guidelines. Also, I doubt they'll follow point 2B3:
> For each result released, the AI lab should make public the name of the model, the prompts used, a (summarized) chain of thought, the time taken, and the estimated cost of computation. Releasing additional material that sheds light on the scientific process that produced the results, such as the initial LLM outputs before they were cleaned up (as recommended in (1) above), is strongly encouraged.
- crypto420
I kind of see their point.
For thousands of years mathematics had this aura of being at the pinnacle of human thought and was one of the things that separated us from other species. People like Euler, Gauss, Euclid, Hilbert, Erdos had this aura about that them that they were on the noblest of pursuits - trying to solve problems that had no immediate practical use, and largely needed focused thinking over extended periods of time with the occasional spark of creativity.
Then comes along AI, a mass extinction event to the entire field. It turns out LLMs can think for longer without stopping, have better recall and working memory. What we're seeing now is a 5000-year old set of institutions facing an existential threat because they no longer require humans to be the sole source of labor.
I find the closest analogy to be automobiles displacing the horse-carriage industry in the early 1900s. Horses couldn't protest and rapidly went out of a job, but horse-carriage owners and that industry certainly did.
https://ghostsofdc.org/2018/07/16/horse-dealers-claim-automo...
- fidotron
"Mathematicians did not ask for this work to be done."
Gatekeeping much?
"Mathematicians have a particular vision of progress that is informed by history and field-specific considerations."
This cannot seriously have been written by anyone mathematically literate, it's just too embarrassing.
- woopwoop
I am a former mathematician now working a regular job (not affiliated with OpenAI). Just speaking for myself, I find this statement deeply embarrassing and at odds with the spirit of open inquiry for which the community of mathematics has (nearly) always been a standard bearer.