This archive copy may be incomplete. The original-source link is preserved in the card footer.
RT @lumpenspace: Not to bang on the usual drum, but: To properly model a human’s semantic and causal network An LLM cannot have to swerve…
More notes on nearby ideas
Quote tweet · Mar 28, 2023oo fun... these sort of semantic relationships trained into the LLM is what I was hoping to get at w the question of building "maps"…Reply · Jun 5, 2024@RokoMijic imo 90% right- you do need interaction w the real world... but LLMs w better prompting & synthetic data & agent wrappers that interact w…Reply · Mar 19, 2025@ESYudkowsky LLMs are experiencing the world through a very narrow channel, just a limited context window of tokens, often mostly their own previous…Quote tweet · Mar 18, 2023same, I'm not surprised that LLMs are human-like but that humans are so LLM-like https://x.com/Scobleizer/status/1637163731510779904