Wonder how a "self-play" equivalent would look like for LLMs, since they have no easy criterion to evaluate how well they are doing like in Go (as mentioned in the videos).
While this would still rely on a knowledge base in the LLM, I would imagine it could simplify the effort required to train reinforcement learning models, while widening the domains it could apply to.
Of course in the real world, for a real intelligent system, reality is the feedback/reward system, but for an LLM limited to it's training set, with nothing to ground it, maybe this is the best you can do ...
The idea is essentially that you need to assume (but of course GI-GO) that most of the training data is factual/reasonable whether in terms of facts or logic, and therefore that anything you can deduce from the training data that is consistent with the majority of the training data should be held as similarly valid (and vice versa).
Of course this critically hinges on the quality of the training data in the first place. Maybe it would work best with differently tagged "tiers" of training data with different levels of presumed authority and reasonableness. Let the better data be used as a proxy for ground truth to "police" the lesser quality data.
1) Really need runtime prediction feedback, not just pretraining
2) Really need feedback on results of one's own (prediction-driven) actions (incl. speech), not just on passive "what will happen next" observations
If this was actually done (and this is just wild baseless speculation), this would be a good reason to let Sam go.