Yes, and really the only feasible way economically to run it is to inference on the edge, that’s to say, in a Web3 model, I have the people participating to run their own generative AI companions on the device. And it’s a simple matter of economy because unless you happen to be Microsoft Bang or Google Bar, you’re not going to be able to afford the kind of inference hardware for millions of users to run language models at the same time.

Keyboard shortcuts

j previous speech k next speech