Safety
that said, another tweet by the same user does correctly assess what’s wrong with most user’s model of what LLMs say.
Gary Marcus critiques common misconceptions about how large language models (LLMs) function, suggesting that most users have an incorrect mental model of what LLMs actually do when generating response
Gary Marcus critiques common misconceptions about how large language models (LLMs) function, suggesting that most users have an incorrect mental model of what LLMs actually do when generating responses. The tweet appears to be part of a broader discussion about the limitations and capabilities of LLMs, correcting a previous claim while establishing a more accurate understanding of LLM behavior.
Related
- ChatGPT doesn’t know its whisk from its elbow
- good to know that a trillion dollars’ investment scaling has thoroughly solved the challenges of common sense.
- Why have many programmers gone back to handcoding? Because with AI (shocker!) it’s….
- LLMs Know They're Wrong and Agree Anyway: The Shared Sycophancy-Lying Circuit
Source: Gary Marcus (X) | 2026-04-27