Turing laureate Rich Sutton says LLMs can vary but can't truly discover. The thread disagrees (twitter.com)
His claim: a model's trajectory is 'either novel or good — based on randomness or based on data — but never both at the same time.'
The rebuttal that stuck: this describes pretraining, not the RL era. Put a model in a loop with tests and feedback and you have evaluation and selection.
Sign in to save