@slowcoding@social.vivaldi.net
2026-09-09 20:47 UTC
@briankung@hachyderm.io @tao@mathstodon.xyz agreed. But as a non mathematician I am curious to understand what is the core difference between what LLMs do and what a more standard theorem prover could do, given the same monstrous compute power?
In other words, is that not just a roll of a gigallion dice’s, a massive brute-force solving with a very energy-inefficient fancy stochastic method, or do we have elements to tell this is more?
Replies (1)
-
@jmopp@masto.ai 2026-09-11 05:45
@slowcoding@social.vivaldi.net @briankung@hachyderm.io @tao@mathstodon.xyz It's the same as the difference between writing a program and using an LLM for coding. The former still requires a human to set up the logic, and to perform the exploration required to come at a solution