Translating KV cache between models

Harry speculates about fast translation of KV caches to let different models share rich context instantly.

Play episode from 34:00

Transcript

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!