← @bgSponge

To grieve, or not to grieve?

Summary

On the Xena blog, xenaproject says language models are solving hard mathematical problems humans could not, and sorts many colleagues’ reactions into denial, anger, bargaining, and depression without treating that map as a criticism.

  1. Many reactions fit the early stages of grief.

    The post illustrates denial, anger, bargaining, and depression with the Association for Human Mathematics, angry essays, the AGMAI recommendations, and colleagues shaken by the Navier–Stokes news, and says this map is not a criticism.

  2. Understanding is an unstable fallback.

    Prize talk is shifting from the hardest theorems to the deepest human understanding, and the author warns that this retreat fails if models soon explain proofs as well as they find them.

    Language models are solving hard problems which humans could not, and mathematicians are reacting very differently to this news.

  3. The author wants a theorem tally.

    They say they never mastered every dependency of their own work, trust Lean because it rejects proof by authority, and are in it for theorems correctly proved.

    I’m in it for the theorem tally.

  4. Publish the withheld results.

    They signed a letter asking OpenAI to dump the significant theorems it says it is holding, calling the secrecy akin to censorship and leaving human reverse-engineering for later.

  5. Infinity is the bound they trust.

    Machines already show evidence at problem solving and little decisive evidence at theory building or conjecture, and infinite mathematics outlasts any finite exponential spend.

Read the original ↗Save to your Sponge

Summarized with Sponge. Quotes are the author’s own words.

More from @bg

All readings →