Recently Dwarkesh Patel shared some of his thoughts about one of the major challenges the current crop of large language models face: they’re not easily trained like a human assistant can be.
… the fundamental problem is that LLMs don’t get better over time the way a human would. The lack of continual learning is a huge huge problem. The LLM baseline at many tasks might be higher than an average human’s. But there’s no way to give a model high level feedback. You’re stuck with the abilities you get out of the box. You can keep messing around with the system prompt. In practice this just doesn’t produce anything even close to the kind of learning and improvement that human employees experience.
Today on X Andrej Karpathy replied to Dwarkesh and included the following which introduced a new term to me describing this weakness:
I like to talk explain it as LLMs are a bit like a coworker with Anterograde amnesia – they don’t consolidate or build long-running knowledge or expertise once training is over and all they have is short-term memory (context window). It’s hard to build relationships … or do work … with this condition.
I’m quite interested to see how this issue begins to be meaningfully addressed!
You must be logged in to post a comment.