can there be GAN LLM?one karen LLM that critiquesone timmy LLM that constantly fixes everythingone jew LLM to evaluate, supervise, tell when to stop or change strategy
Yeah, you can totally do that.
>>108570749I think traditional GAN is kinda inefficient because you want to have some kind of runtime recursivity in your model.
>>108570749You can do everything you want, usually these kind of multiagentic workflows work way worse than imagined, because LLMs tend to develop schizophrenia over time.
>>108570749You can do this now
>>108570749GAN is hard to train
Claude has agent team mode exactly for this purpose.
>>108570795what you mean you can always turn those things into flat loop
