4 ms·
We can probably make it work at more nebulous problems by using a big LLM to judge the quality of the answers as well. It should be easier to recognise e.g. a r
by vintermann 2y ago
We can probably make it work at more nebulous problems by using a big LLM to judge the quality of the answers as well. It should be easier to recognise e.g. a really good poetic translation than to make one, and as long as that's true it could benefit from internal monologue "reasoning" in those domains as well.
- timr 2y agoThis has been done, and doesn't work especially well. See, for example, GANs, which are difficult to train and vulnerable to mode collapse. Not saying that it's impossible, but reinforcement learning of the form shown by DeepSeek was particularly well-established and robust. It's sort of ironic, actually...IIRC, OpenAI started in the world of this kind of reinforcement learning.