3 ms·
This is not how ML works generally nor LLMs like ChatGPT specifically. What you're describing sounds like RLHF, which changes the style of responses and impact
by MacsHeadroom 3y ago
This is not how ML works generally nor LLMs like ChatGPT specifically.
What you're describing sounds like RLHF, which changes the style of responses and impacts things like refusals but does not add to a model's intelligence (in fact it reduces model intelligence).
An LLM's intelligence comes from pretraining in which there are no prompts, or answers, only corpus and perplexity.
- mjburgess 3y agoYou can map the weights to an uncompressed (Q,A) space, and then map them back again to weight-space -- all without loss of information. The actual domain space they were compressed from is irrelevant; their values are equivalent.
- MacsHeadroom 3y agoQ & A are not part of pretraining.