AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Importance of Pre-Training Everything
The examples in the live launch stream that they did of understanding the images definitely blew my mind. They have some really interesting techniques which I imagine, you know, open eyes doing some similar stuff. Their approach is involves training a connector model to essentially translate an image in coding to the latent space of the text model. It's actually predicting the embeddings and kind of injecting them directly into context.