We posit that the superior writing abilities of LLMs, as manifested in surpassing human annotators in certain tasks, are fundamentally driven by RLHF. — Llama 2
I think we've seen exactly what META can do when it chooses to throw a rock into the pond of social media companies. Threads was adopted stupefyingly fast not just because Twitter is a toxic waste dump, but also because they have a ridiculous user base.
At some point, LLM training is a numbers game, and META is absolutely in the game (for now, anwyay) in a way nobody else was really threatening Open AI and Microsoft's supremacy.
Regarding RLHF, I love thinking about 'human preferance.' It's like my cycling routes – what really guids those complex reward functions?
I think we've seen exactly what META can do when it chooses to throw a rock into the pond of social media companies. Threads was adopted stupefyingly fast not just because Twitter is a toxic waste dump, but also because they have a ridiculous user base.
At some point, LLM training is a numbers game, and META is absolutely in the game (for now, anwyay) in a way nobody else was really threatening Open AI and Microsoft's supremacy.