9 Comments
User's avatar
James Whelan's avatar

The vision of inheritantly unstable magnetic fields being controlled by AI in vehicles with dire consequences if there is a glitch is not reassuring.

I like the jump to magnetic induction to produce 'electricity' rather than the continuation of the kettle, steam, lump of rotating iron methodology.

Karl Sanchez's avatar

Yes, there clearly are some very dangerous aspects of energy generation via fusion regardless the method. AI, IMO, was designed to deal with the types of computations fusion production requires, not to provide answers to basic questions. That feedback won’t work with fusion means “feedforward” needed to be developed using AI to rapidly do all the computations. He described the process of writing the program to deal with that problem. He said: “Based on solvers, we have built China’s first FRC balanced AI agent model based on encoding-decoder neural networks.” and provides further description.

Richard Roskell's avatar

Fascinating stuff, thanks Karl. Very cool that the FRC reactor produces electricity directly by induction, rather than going through intermediate steps.

Karl Sanchez's avatar

Yes! Boiling water to generate steam to rotate turbines is so 19th Century. The idea that a fusion reactor can be scaled down is also revolutionary. The US effort by Helion is North of me at Everett in Washington state.

Diana van Eyk's avatar

This is pretty exciting stuff, Karl. Nuclear fusion is getting close to becoming a reality. Thanks for letting us know about this.

Natasha's avatar

Global supplies of civilian-use tritium are estimated to be approximately 25 kg, with most of it produced by Canada. Tritium fuel is one of the weakest part of the fusion supply chain. The other is maintaining hundreds of millions of kelvin for more than a fraction of a second. Even if these obstacle can be overcome, modest‑performance breeders (TBR≈1.05–1.2) produce only small surpluses (grams–sub‑kg per plant per year), so existing ~25 kg stock funds can only supply ~5 start up plants, meaning fleet growth is slow; high‑breeding designs (TBR≫1.5) could generate multiple kg/yr per plant and enable faster expansion (several new starters per year per tens of operating plants).

WTFUD's avatar

Way out of my league Karl but my A-level physics mentioned Theoretical v Practical, Mechanics v Applied Mechanics.

I'll maybe give it another go tomorrow.

Karl Sanchez's avatar

I suggest focusing on the business applications and associated involvement of China’s technical and government resources. Plasma physics is where it’s at along with understanding of how AI can be integrated for optimal results. Compare what you read with that old Star Trek/Star Wars stuff. Just wait for the harnessing of gravity—The Force.

Mark Watson's avatar

Totally sceptical. IMO fusion reactor boom is the next boom after the AI boom which is on the skids . Note that AI companies are building and restarting old fission reactors to power their data centres . I see speculation but no working reactor still . Watch the money get poured into this same as AI.