3 ms·
ChatGPT and Syndey were extensively trained with chat specific training via RLHF (realtime learning from human feedback), which is missing from GPT-3.
by LesZedCB 4y ago
ChatGPT and Syndey were extensively trained with chat specific training via RLHF (realtime learning from human feedback), which is missing from GPT-3.
- warkdarrior 4y agoRLHF = _reinforcement_ learning from human feedback
- visarga 4y agoHaha, someone was asking today on Twitter > I can't find a mention of "GPT" or "General Purpose Transformer" :)