sillysaurusx 4 hours ago

It’s been said that RL is the worst way to train a model, except for all the others. Many prominent scientists seem to doubt that this is how we’ll be training cutting edge models in a decade. I agree, and I encourage you to try to think of alternative paradigms as you go through this course.

If that seems unlikely, remember that image generation didn’t take off till diffusion models, and GPTs didn’t take off till RLHF. If you’ve been around long enough it’ll seem obvious that this isn’t the final step. The challenge for you is, find the one that’s better.

  • whatshisface 3 hours ago

    RL is barely even a training method, its more of a dataset generation method.

    • theOGognf 2 hours ago

      I feel like both this comment and the parent comment highlight how RL has been going through a cycle of misunderstanding recently from another one of its popularity booms due to being used to train LLMs

      • mistercheph 5 minutes ago

        care to correct the misunderstanding?

  • paswut 2 hours ago

    What about for combinatorial optimization? When you have a simulation of the world what other paradigms are fitting

  • charcircuit an hour ago

    GPT wouldn't have even been possible, let alone take off, without self supervised learning.

zerosizedweasle 4 hours ago

Given Ilya's podcast this is an interesting title.