“Humans are generative models. And there is a lot of innovation in what they are calling prompt engineering and all kinds of things but the structure of it is almost like AI is not philosophical enough yet to be “thinking”.
Humans think in terms of overall goals, moral standards, stuff our parents told us to do, short term goals, long term goals, constraints by our friends and society, etc and all of these are present when we are doing our daily tasks. And this is mostly not instantaneous inference and mostly not training.
And then the fact that these big generative language models are also starting to do that is really curious. And I’m thinking a lot about how you would build a computer to do that better.”
Listen to Tenstorrent CEO Jim Keller talk about building technology and teams on the Computer Architecture Podcast here.