Just Too Alien?

Here’s a boot to the shins:

More than 1000 mathematicians have pledged not to work for AI companies and to avoid using their models – to varying extents – in their research. The move comes amid a growing backlash against the way AI companies have used their models to solve thorny mathematical problems, many of which had stumped humans for decades. …

Terence Tao at the University of California, Los Angeles, said last month that the way in which these discoveries are being made and “dumped” on the mathematical community may harm the field. While in many cases the results are formalised, so can be relied upon as accurate solutions, they are often presented in an impenetrable and hastily written way that leaves the field struggling to extract meaning and understanding. [“1000 human mathematicians pledge to shun AI in their research,” Matthew Sparkes, NewScientist (31 October 2026, paywall)]

Rejection by a group that might have been considered potentially one of the most welcoming can be a bad sign for artificial intelligence. For a quick background, if a proof of some assertion is “formalized,” meaning translated into a computer language (I’m not sure that’s the right term, but OK) named Lean, then it can be inspected and verified automatically. However, while the existence of a proof is useful, the techniques learned while building a new proof may be more valuable than the proof itself.

Apparently, mathematicians are deprived of those techniques in AI-generated proofs. They’re simply too alien to extract, much less understand.

This isn’t pouting, but legitimate complaints in my view. I’ll be interested to see if the effort falls apart, or if more mathematicians join.

Bookmark the permalink.

About Hue White

Former BBS operator; software engineer; cat lackey.

Leave a Reply

Your email address will not be published. Required fields are marked *