←back to thread

Symbolic Logic Based LLM

5 points sandeeptshelvan | 1 comments | 15 May 25 23:35 UTC | HN request time: 0.221s | source

Why big AI labs not building LLM that fully trained on Symbolic logic instead of Natural language?
1. NoahZuniga ◴[18 May 25 03:52 UTC] No.44018887[source]▶
Because most useful things in our real world aren't expressed in symbolic logic. Also Google with alpha proof has shown that LLMs trained on Natural language can produce pretty good symbolic logic as output.