This Day in AI Podcast cover image

Nvidia & The AI Goldrush, Neuralink Human Trials, LLM Specialization & Meta's MEGABYTE | E16

This Day in AI Podcast

00:00

The Benefits of Facebook's Open Source Platform

Instead of using individual tokens, they're using these things they call patches. So when we talk about a token, we mean like two or three characters. And the idea is the models are trying to predict the next token. What this does is takes a couple of words, like one or two words or maybe three words and use patches of them. Then divides those up and uses smaller models within that to analyze them. It can have a much, much like a universe of knowledge like you're saying within it. The task could range from booking an entire vacation, hosting a conference could be as simple as talking to the large language model.

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