4 ms·
The other element here is quality content - you probably aren’t just training on public internet data for commercial LLMs, hopefully you can train on scanned bo
by Closi 4y ago
The other element here is quality content - you probably aren’t just training on public internet data for commercial LLMs, hopefully you can train on scanned books too, and closed academic journals, radio transcripts, photograph stores, map data, codebases, technical documentation...
- lostpharoah 4y agoI suspect a lot of low-quality content still makes it into the training process and the RLHF layer is specifically designed to improve the quality of the output despite noisy input. i.e. without RLHF, these models would for sure be vulgar, explicit, etc.