Ask HN: Did Google know about RLHF(breakthru) only after OpenAI shared

[elarocks]

I am still unable to think of a reason why Google, Anthropic and AWS did not know about/invest in RLHF, before OpenAI shared their success around implementing and in scale that was viable. Would you say that if OpenAI had not shared about RLHF, Google and Anthropic wouldn't be where they are today ?

[bigyabai]

RLHF is basically a fancy, overengineered GAN. Most of the industry could see that DPO was more efficient for fitting to human behavior.