139 - Coherent Long Story Generation, with Kevin Yang
NLP Highlights
The Problem of Coherence in Long-Term Generation
In general, can you give us a sense of what people have tried previously to address the problem of coherence in long-term generation? Yeah, yeah. So first of all, let me say like prior work for the most part focuses on like much shorter generations. And my understanding is this is because once you scale up to GBT3 like Da Vinci, like a lot of the problems on the scale of like a few sentences are largely solved. But when you're using like the older models like prior to this era of like pre-training or like massive scale pre-training, then then you actually have to worry about a lot of problems like repetition and not making sense at all
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
