I should say, therefore ... GPT-3 (the model you use for text) is not *remotely* the same as Dall.E (the model you use for images) but they were both created using the same kind of underlying technol… (View Post)

"it is not true to say they have no relationship" is correct. "They are NOT two entirely different systems" is only correct if "entirely" is taken to mean "share no… (View Post)

In fact, you are both sort of right. The thing that underlies both Dall.E and GPT-X is a domain agnostic transformer. When it is trained on a big old set of data, it produces a model (e.g. Dall.E or … (View Post)

When she calls you "one of her playschool babies" you *think* you should cringe but actually you melt. She'd sort us all out. (View Post)

I agree that both types of commentator in the wider debate are wildly wrong. I would classify all of them as the interesting voices. And yes, you have to apply Clarke's first law to people like Yann … (View Post)