A researcher claims Microsoft and OpenAI may have cracked multi-datacenter distributed training for their AI models based on their 'actions': "Mic...

Roccy

New member
May 4, 2017
4
1
3
Visit site
It will be interesting to see how this power-hungry tech fares with the looming degradation of the US power grid as green power takes over from the current grid's more reliable sources that are already starting to be shut down.
 
  • Like
Reactions: nop

fjtorres5591

Active member
May 16, 2023
410
108
43
Visit site
It will be interesting to see how this power-hungry tech fares with the looming degradation of the US power grid as green power takes over from the current grid's more reliable sources that are already starting to be shut down.
Well, MS is looking at nuclear (reactivating Three Mile Island according to some reports), fusion (the deal with Helion for 2028) and lots of co-located solar. They're placing bets on lots of potential solutions.
Google is betting on geothermal fracking.
The others?
Unclear.
 

nocturn9x

New member
Dec 27, 2023
5
0
1
Visit site
Well, MS is looking at nuclear (reactivating Three Mile Island according to some reports), fusion (the deal with Helion for 2028) and lots of co-located solar. They're placing bets on lots of potential solutions.
Google is betting on geothermal fracking.
The others?
Unclear.
Honestly, the more I look into Helion the more it seems like it's a lot of marketing/PR nonsense. But I'm in no way a fusion expert so I guess I'll just see what happens
 

fjtorres5591

Active member
May 16, 2023
410
108
43
Visit site
Honestly, the more I look into Helion the more it seems like it's a lot of marketing/PR nonsense. But I'm in no way a fusion expert so I guess I'll just see what happens
The thing about Helion is their approach to fusion isn't the miniature star approach of the tokamaks but more of a first principles approach. Think airplanes and helicopters versus ornithopters.

Instead of trying to maintain a mini star, super hot plasma stable long enough to extract useful heat, the Helion system uses pulsing electromagnetic forces to collide plasma clouds to generate energy and extract the generated energy. (rather like an electromagnetic diesel-style engine). Its not reliant on thermodynamic cycles or long containment times but on many tiny fusion "explosions" in sequence.

Much like the Farnsworth fusors and Bussard Polywell wiffle balls, their approach already works, but haven't yet produced enough energy to be useful. They need to scale up their device pulse rates successfully to produce more energy than they consume rather than grow physically bigger or increase containment times. To date their prototypes have scaled according to theory. They may or not succeed but they haven't failed yet.

The PR hype is actually necessary because, remember, it is a private tech startup rather than a multinational jobs program like ITER or a government lab project. They need to lure investor support them with the promise of big profits and in today's economic environment venture capital is harder to raise for anything other than "AI".

Whatever they have, they convinced to sign up for *production* output, with economic penalties if they don't deliver.

 
Last edited:

Members online

Forum statistics

Threads
327,049
Messages
2,249,282
Members
428,592
Latest member
Hanik