Page 1 of 1
Deepmind AI plasma control
Posted: Sun Feb 20, 2022 4:00 pm
by RERT
Quite a lot of coverage of this. Here is an article from Nature....
https://www.nature.com/articles/s41586-021-04301-9
And some skeptical comment (Not sure I'm convinced of this one).
https://wattsupwiththat.com/2022/02/19/ ... n-plasmas/
Re: Deepmind AI plasma control
Posted: Mon Feb 21, 2022 3:07 pm
by Giorgio
That was a darn interesting read and a great job from the team.
The skeptical comments are out of context as they do not focus on the actual results but on the underlying architecture that was used to reach those results. He might also have a point, but I do not see how this can invalidate the work presented in the paper that (in my personal opinion) has huge implications for any fusion confinement technology being developed right now.
Really a great result.
Re: Deepmind AI plasma control
Posted: Mon Feb 21, 2022 11:09 pm
by RERT
No, I agree, the sceptic felt wrong for two reasons: 1) A stateless model might include ‘rates of change’ as inputs to avoid itself computing state changes and 2) his position is actually optimistic, in that his contention is that a better architected AI would do an even better job.