A new AI system from Google adds a skill that Chat GPT lacks: logical reasoning. It recently outscored most human entrants on Intl. Math Olympiad problems, demonstrating that it can construct lengthy logical arguments that hold up to scrutiny.
Large Language Models like Chat GPT can generate gushers of plausible explanations but can’t tell if they’re right or wrong. Adding the new reasoning capabilities to the existing generative ones could be the next big step forward.
As usual, Sabine Hossenfelder explains it better than I could:
Steven Chalem's Blog 
Leave a comment