At least for quantum I think I prefer ISC. Much easier to talk to people.

All in all, IMHO, Jensen statement is in a superposition of negative and hype - depending on what you mean by very useful.

It's not crazy to think QPUs will have a similar path, where some narrow, unreliable, use-case can be obtained. In 20 years those might be less narrow. But probably still a very long way from being ubiquitous — just like GPUs…

Also, read some reports on how difficult it is to run large GPU clusters over time. In fact critical systems on e.g. boats, last time i checked, cannot use GPUs as they won't get the required reliability ratings. None of this has stopped Nvidia’s GPUs finding various “killer apps” over 2 decades.

In 2019-2021 I was working in computer vision applied to videos. We tried to use deep learning to extract real-time information from video, all on GPUs. The cost of GPUs, and DL problems, completely killed that field. From what I see people have largely moved away from the approach at the time.

What are “very useful #quantum computers”?

Much has been said about Jensen's latest comment of 20 years until that date. One discussion I've not sufficiently seen so far is discussing what we call “very useful”.