[Levitating]

I should clarify, I believe language models will have an enormous technological impact. But the media and companies with a stake in it are blowing it way out of proportion. Like nvidia saying that they will create AI games. And it has caused people to extremely overestimate what a language model can do.

There is no reason to believe OpenAI will magically change the phone industry by making "OpenAI phones" whatever that is supposed to mean.

AI assistants in phones were a flop, and though their quality could change much with the use of language models this doesn't mean that people are going to use them this time. Currently it's not even possible to use language models in that sense, they are useful for generating text but it's near impossible to have them communicate to some external interface.

I've once talked to someone who gave a talk on an absolutely enormous call support framework for a bank, and he told me that they had no plans to ever integrate actual AI with their chatbot. Because those models can't be controlled/consored strictly enough.

What I am saying is that the actual technological advancement of language models has completely disassociated with the buzzword "AI" that is being thrown around a lot right now. And it's not dissimilar to how Meta or Web3 were discussed.