Tech's hype artists didn't predict ChatGPT either. Hype artists predict the ambitious futures found in sci-fi novels, but these futures are not what actually occur.

Only the people actually inventing the future, things like AlphaGo and GPT, have a limited crystal ball into the future.

Tech pundits on LinkedIn and Twitter and HN have rarely if ever gotten anything right about what exactly the future will look like, and I expect that pattern to continue.

To be fair though, I still don’t think ChatGPT is the AI we expected to be talking too. The talking AI was a Hal 9000, current an LLM isn’t that and it might never be. Time will tell I guess.

Personally I think there is a hidden special property of ChatGPT-4 that makes it so unreal. It talks and because it talks, we can say this: “I bet you never thought that would happen, we’re close to solving all problems.”.

Not discounting the talking bit but if it washed dishes or even those that just generate images, we’d be less convinced the AI is about to take over. Someone once commented, why don’t we worry about stable diffusion taking over the world. I have to say I do agree a little because being trained on image data must help it build some type of world model too. We just don’t worry about that though.