Maybe. It kind of reminds me of how modern fighters can be unstable (to give them better performance) through the use of smart avionics.
18 thoughts on “A Fusion Breakthrough?”
Comments are closed.
Maybe. It kind of reminds me of how modern fighters can be unstable (to give them better performance) through the use of smart avionics.
Comments are closed.
Hopefully the AI won’t be crippled by diversity “guiderails”.
Baryonic matter half-lifes matter.
Going into Dwight Mode:
Unlike nuclear fission—the nuclear reaction that is currently used in the energy sector—fusion does not create radioactive waste.
Wrong!
The approach in question involves using artificial intelligence.
I’m not holding my breath, Jim!
Reverting to Dave Mode:
Should be ready in about 20 years.
Was House Speaker Mike Johnson in “Dwight Mode” last night?
I so wanted the Speaker, every time Biden mentioned a new spending program, to pull out a wad of dollar bills and start tossing them off the rostrum. Take that paper show Nancy…
These gushy articles tend to gloss over nasty little details as to what the fuel is, in this case almost certainly neutron-rich DT plasma. So, although not as nasty as solid waste fuel rods from LWR reactors, the containment vessels will get pretty toasty. In any case, this story is a couple of weeks old. I’m not sure that anyone is going to put their own giga bucks into something that relies on LLM software that that tends to say that Vikings were black.
“I’m not sure that anyone is going to put their own giga bucks into something that relies on LLM software that that tends to say that Vikings were black.”
The bizarre results were to some extent a result of corrupting the **input prompts** with wokeism, it is unclear as to what extent the LLM proper is corrupted.
It seems possible that, for a price, government and large corporate actors could get access to the LLM with clean prompts.
Yes, the containment vessels will become low-level waste that have to be managed for decades, not centuries.
Unlike nuclear fission—the nuclear reaction that is currently used in the energy sector—fusion does not create radioactive waste.
Friends of mine who are nuclear engineers oppose fusion because of its prodigious radioactive waste output. Economical fusion will require deuterium and tritium fuel; stable deuterium is available in abundance on Earth, while tritium, with its 12 year half-life, must be manufactured. That’s most readily done by bombarding lithium 6 with neutrons, though the neutron economy of fusion doesn’t close by itself. There has to be neutron multiplication, involving beryllium to generate additional neutrons by fast neutron-beryllium reactions, and alpha-beryllium reactions from the high-speed fusion products. You have to over-produce secondary neutrons to assure an adequate supply of tritium, and the ones that are left over will activate any other material left in the reactor. Ultimately, a commercial power plant will be a radioactive hot mess.
And don’t think the so-called “aneutronic” deuterium-helium3 reaction will be a panacea. There isn’t even a theoretical path to making it happen, because of the massive bleed off of plasma energy via bremsstrahlung radiation. Further, there is no source of helium 3 on earth adequate to supply our energy needs. It can be manufactured…by making tritium from lithium and neutrons, and collecting the decay product, which is helium 3. Not a big improvement.
I think the plasma stabilization work that’s been done is really first-rate, and the AI component is a significant step forward. But the engineering barriers yet to be overcome after demonstration of a stable burn of arbitrary duration are massive. At the level of research expenditure on fusion, it will always be 30 years away.
My understanding is there is plenty of Helium 3 on the moon and my understanding of a radioactive containment vessel/parts is they are only something that needs to be dealt with for around a century before it’s cooled to background.
Right!
Its interesting that the AI is essentially doing what SpaceX is in iterative design, fail, succeed testing. Only on a much compressed timescale.
Only on a much compressed timescale…
Twenty years from now instead of sixty (60) years from now…?
Or as the Q conspiracy folks say on the chans “two more weeks”. 🙂
A neural net.
Except that the AI is doing iterative modelling. No doubt SpaceX is also modelling, but they are actually building stuff (and having it blow up occasionally).
The AI will be limited in its accuracy unless they actually build stuff at some point and have the risk of it failing.
AI is so hot right now.
If you want More funding for your field of study, link it to something that’s already high priority and politically correct.
AI:Fusion:AGW.
And it will prevent gullible warming.