News flash: MATH ON MARS? World's richest billionaire raids Google Research…

…relegating mechanization of mathematics to a footnote.

Szegedy and Tony Wu, whom readers of Silicon Reckoner have previously encountered, touring SpaceX Starbase in Texas

                               Math for AI and AI for Math! (writes Greg Yang)

All I know is what I’ve read in the papers, and on Twitter, where we learn that

X.AI will be a 'pro-humanity' artificial superintelligence that would be maximally curious about humanity rather than having moral guidelines programmed in it.

because Elon M. (whom readers may remember for “predicting artificial general intelligence by 2029, and ‘Hopefully, people on Mars too’”),

believes that by nurturing a strong sense of curiosity within the AI, its behavior will naturally align with human values and mitigate potential risks associated with the development of AI.

This is huge news — two days after the announcement a Google News search for xAI promises “about 58,900 results,” many of them in the business press. It’s probably good news that mathematics is absent from all but a handful of them. These quote Greg Yang (formerly of Microsoft Research) to the effect that

the startup will delve into the “mathematics of deep learning,” a facet of AI, and “develop the ‘theory of everything’ for large neural networks” to take AI “to the next level.”

The header above, which celebrates future mathematical musketeers, is copied from Yang’s Twitter page. Here is Yang’s complete tweet:

The mathematics of deep learning is profound, beautiful, and unreasonably effective. Developing the "theory of everything" for large neural networks¹ will be central to taking AI to the next level. Conversely, this AI will enable everyone to understand our mathematical universe in ways unimaginable before.

Math for AI and AI for math!

Any mathematician/theorist excited about this needs to DM me!

If your excitement leaves you time to drop me a line, I’m sure readers of Silicon Reckoner are “maximally curious” about the starting salaries this new rival to ChatGPT etc. that promises “to have your minds blown” will be offering. Anonymity guaranteed!

1. This seems to be an allusion to the interpretability problem, which came up several times at The Workshop. For example, in my report on Alhussein Fawzi’s presentation I wrote that

Fawzi thought quantifying interpretability might be a plausible quantitative benchmark for progress; in the same vein, he also thought mathematicians’ insights might help make AI more interpretable.

Math for AI, in any case.