3 ms·
GPT is too simple. The problem probably requires reinforcement learning, where the AI is trained by interacting with a real computer system.
by amelius 4y ago
GPT is too simple. The problem probably requires reinforcement learning, where the AI is trained by interacting with a real computer system.
- mountainriver 4y agoChatGPT uses RL actually, but you are talking about something more like Adept
- amelius 4y agoThis is new for me. How does ChatGPT use RL? I'm talking about RL in the sense of interacting with an environment during training. Isn't GPT trained by predicting the next word in a large but static piece of text?
- amelius 4y agoUpdate: ok, I found an explanation here: https://huggingface.co/blog/rlhf https://huggingface.co/blog/rlhf