The 2022 AI Index
The Data Exchange with Ben Lorica
00:00
Using Multimodal Models to Generate Images and Text
In recent years you've seen models like clip and others emerge where youare training on both text and images jointly. It's a really exciting area, and progress in it is going to be slower because it's a bit more com heavy. Once i'm doing multiple modalities, te compute budget goes up, like, quite significantly. But imagine how crazy that's going to be when that arrives.
Transcript
Play full episode
