The Sentience Institute Podcast cover image

Raphaël Millière on large language models

The Sentience Institute Podcast

00:00

The Relationship Between Images and Referential Competence

DALI2 learns by being trained on pairs of captions and images. But it's not actually getting any feedback about what words refer to out there in the world. By contrast, chat GPT when it's fine-tuned with explicit human feedback might be achieving some limited form of referential competence just by virtue of getting direct human feedback.

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app