People need to have their expectations set on ChatGPT, just like how they have to set their expectations on Dall-E/MJ/etc. My experiences so far with both as a

People need to have their expectations set on ChatGPT, just like how they have to set their expectations on Dall-E/MJ/etc. My experiences so far with both as an ML person with interests in both domains is that they get things 80-90% right, with some big flaws. Just like how Dall-E can paint a human face pretty well but then mess up on their fingers and other extremities, ChatGPT can give you some really logical-sounding statements but totally bomb the actual facts of things. Am curious how they'll execute - I def see a lot of potential usefulness.