[HN Gopher] Avoiding fusion plasma tearing instability with deep...
___________________________________________________________________
Avoiding fusion plasma tearing instability with deep reinforcement
learning
Author : karma_daemon
Score : 71 points
Date : 2024-03-02 14:44 UTC (8 hours ago)
(HTM) web link (www.nature.com)
(TXT) w3m dump (www.nature.com)
| seydor wrote:
| How about we add transformers to the reactor? But not that kind
| of transformer
| baq wrote:
| Turns out paying attention to plasma is all you need.
| api wrote:
| I remember having this idea when I was studying machine learning
| in college. I'm _really_ happy to see that it occurred to someone
| else in a position to actually look into it, because it "felt
| like something might be there" to me.
|
| The basic idea I had was that fusion plasma containment involves
| containing a turbulent, dynamical system, so it might require
| some kind of actual intelligence learning or co-evolving with the
| system.
|
| I wondered if this might be the only way to achieve over-unity
| fusion outside gravitational confinement (stars, black hole
| accretion disks, etc.). This would mean there are two fusion
| mechanisms in nature: gravitational confinement and cognitive
| confinement. The latter can only be a product of a living system.
|
| When a living system achieves this, its biosphere "ignites" and
| becomes something I termed a "biostar." Biostars could be
| potential SETI targets -- biospheres that have harnessed fusion
| and so emit anomalous amounts of optical and infrared radiation
| on their night side. This moment of ignition would be an event in
| a biosphere comparable to the evolution of photosynthesis-- a
| fundamental change in the energetic dynamics of life.
|
| In the far future life the that achieved fusion could settle
| things like rogue planets in deep space, so that would be another
| potential SETI target. Find objects emitting anomalous infrared
| in the interstellar void. The advantage would be being far from
| destructive events like solar storms.
| baq wrote:
| Assuming fusion drives are a thing we should be able to see
| them as they'd be glowing rather bright.
|
| https://www.projectrho.com/public_html/rocket/torchships.php
| gs17 wrote:
| Unless, of course, all the ships are coming straight towards
| us.
| tsimionescu wrote:
| Fusion drives wouldn't allow these ships to get anywhere
| near the speed of light, so we would have ample time to see
| the light they emit long before they arrive.
| piotrkaminski wrote:
| That would actually make them easier to spot, as they'd
| need to flip around for a breaking burn roughly equal to
| their acceleration burn, pointing their engines straight at
| us. (Assuming they want to stop by and say hi, of course.
| If not, then there's not much to worry about.)
| gpderetta wrote:
| > not much to worry about.
|
| Unless they are literally coming straight ahead at us!
| piotrkaminski wrote:
| If they intend to go straight through Earth at
| relativistic speeds (to establish a hyperspace bypass,
| perhaps?), then there's really not much we can do about
| it anyway. :p
| jetbooster wrote:
| In Peter F Hamilton's Commonwealth series of books, true
| artificial general intelligence 'woke up' from the computers
| designed to handle the incredibly complex calculations required
| to make and maintain long-distance wormhole connections. Quite
| analogous to this situation I feel
| SubiculumCode wrote:
| Sounds like the basis of a new Star Trek episode.
|
| Captain's Log: Since we came to orbit Venuuil III to host talks
| between the Klingons and the Venullians, there has been a
| increased incidence of unpredictable fluctuations of plasmas in
| the warp drive containment field. We are now devoting all
| available power to increasing our computer's ability to track
| and predict this rapidly changing phenomenon. Geordi reports
| that at the current growth factor we can maintain containment
| for 22 hours, 47 minutes, 17 seconds. To support the peace
| talks, we will remain here as long as possible.
|
| Captain's Log: Intriguingly, the fluctuations are beginning to
| reveal an embedded temporal distortion that exhibits language
| like patterns. Data has begun working on an interface between
| the containment field and his positronic neural net.
| tekla wrote:
| > This would mean there are two fusion mechanisms in nature:
| gravitational confinement and cognitive confinement
|
| Da fuck? The mechanism is compression. This is like calling
| "generating electricity via a generator", "cognitive energy".
|
| What is ITER? Black Magic that happens to cause Fusion?
| api wrote:
| Fusion is reasonably easy. You can do it in your garage with
| an electrostatic confinement fusor. I'm talking about fusion
| that generates significantly more power than it takes to run
| the reactor. Only that kind is useful as an energy source.
| That so far has been elusive.
|
| Inertial confinement has sort of achieved this but only on
| paper. If you tally up the total input to set up and run the
| system it's still way in the red.
|
| ITER has the potential to run just a bit over unity but it's
| really just a research platform.
| londons_explore wrote:
| While deep learning might be a good control strategy for such
| systems, I very much doubt it is the only control strategy that
| works.
|
| Control strategies in nonlinear systems are an effectively huge
| search space, and deep learning is just one way to find a good-
| ish solution faster.
| AtlasBarfed wrote:
| Except that fusion on earth will likely never be cheaper than
| solar/wind.
|
| I mean that is a cool scifi story, but economics seems to hate
| cool things.
|
| There's this "big lie" that fusion people imply that it will be
| cheap, clean, and limitless.
|
| Cheap is doubtful, clean is undermined by the reality that fast
| neutrons from fusion degrade the reactor to radioactive
| isotopes, and ok the fuel is pretty much limitless
|
| Now, if we can get scalable fusion as viable load levelling, to
| develop it to the point it can be used in space then that's
| some real scifi.
| thechao wrote:
| Fusion will be necessary if we plan to colonize the solar
| system beyond Mars, and not at Jupiter. It's cold, dark, and
| scary, out there.
| pfdietz wrote:
| Power can be beamed out to interstellar distances, so
| fusion isn't necessary.
|
| For that matter, if a space colony is equipped with a
| mirror for concentrating the sunlight needed to illuminate
| the inside as if it were Earth, and we place the limiting
| distance as that at which the mass of the mirror is equal
| to the mass of the space colony, the distance is about 1
| light year.
| tsimionescu wrote:
| We won't, not with anything resembling current technology.
| So, if we were to imagine a colonized solar system, there
| is a good chance it's not fusion that gets us there, but
| some currently unknown technology.
| audunw wrote:
| I thought so too. It's pretty simple: if you're making a
| nuclear thermal power plant a lot of the costs are associated
| with building the containment vessel and the heat exchange
| mechanism. Fusion is fundamentally lower power density than
| fission, so you'll need a bigger vessel for a given power,
| and both the containment and heat exchange mechanism are far
| more complex and expensive. Thermal fusion will never be
| cheaper than fission, which already has a hard time competing
| with renewables. And renewables are still getting cheaper.
|
| But then there's Helion. If you can extract electrical power
| directly rather than through heat exchange and a turbine, it
| changes the equation drastically. So I think their approach
| can work from a theoretical point of view.
| tsimionescu wrote:
| Currently imaginable fusion power plants generate nowhere near
| enough power for the excess to be visible from outer space.
| They would not even be a blip compared to the largest already
| existing hydro power plant, for example.
|
| Edit: to add some numbers, the "planned" DEMO power plant (the
| hypothetical successor of a successful ITER experiment) would
| produce something like 750MW, while Three Gorges Dam produces
| 22,500MW. Even if DEMO could be scaled up (which is hard, given
| that it would already be beyond the limits of today's material
| science), it definitely couldn't scale up 30 times.
| leereeves wrote:
| What stops fusion from scaling up by building more plants
| instead of larger plants?
| tsimionescu wrote:
| Nothing except the costs, but then it's not a single
| reactor anymore and other power plants scale as or more
| easily. So nothing is really special about fusion power if
| we just want to scale horizontally.
| leereeves wrote:
| I thought fusion is special because the inputs are more
| abundant and the outputs less toxic than other fuel power
| plants.
| eganist wrote:
| The concept of "cognitive confinement" (your term I presume.
| It's neat, I hope it sticks) was explored in Spiderman 2.
|
| I'm being entirely serious. Even though the visualization of
| the reactor was wildly off the mark, this specific concept fits
| what the writers had in mind with the AI mechanical limbs.
| fancyfredbot wrote:
| Deepmind worked on reinforcement learning for plasma control back
| in 2022 and this also led to a paper in nature. I don't really
| understand the differences between their earlier work and this
| paper but deepmind don't seem to be involved in this one:
| https://deepmind.google/discover/blog/accelerating-fusion-sc...
| jeffbee wrote:
| The DeepMind paper in turn cites the authors of this paper,
| previously. One of the big differences in the current paper is
| that the experimental device is much larger and more powerful,
| and the duration of the shot is longer as well.
| londons_explore wrote:
| > nuclear fusion is rapidly emerging
|
| And has been rapidly emerging for 60 years...
| aqme28 wrote:
| Same with AI, and yet they both keep getting closer
| klipt wrote:
| Is it economically competitive with solar though?
| exe34 wrote:
| At night it is.
| pfdietz wrote:
| Not with LFP cells falling to $54/kWh.
| aqme28 wrote:
| Is it? Definitely not. Could it be? Maybe. Who knows
| jeffbee wrote:
| Solar power _is_ gravitational confinement fusion.
| londons_explore wrote:
| I kinda get the feeling there are a core of scientists who
| want to work on this stuff.
|
| They'll work on it even if there is no hope of commercial
| fusion power, simply for academic kudos. Government's will
| fund it simply because there isn't much other blue skies
| physics research to fund.
| ben_w wrote:
| Also like AI, we have natural examples as existence proofs it
| is possible, followed by narrow non-generalised artificial
| examples, yet actually getting it working _properly_ is very
| hard.
| SkyMarshal wrote:
| Another team from Japan and the US published a paper on this
| recently too:
|
| https://www.nature.com/articles/s41598-023-49432-3
|
| Sabine has a good review of it:
|
| https://www.youtube.com/watch?v=4VD_DLPQJBU
| devaiops9001 wrote:
| Tokamak is dumb. I'm sick of hearing about Tokamak.
|
| The plasma in the Safire reactor has self-containing magnetic
| fields and doesn't need the $20billion+ super-magnet
| infrastructure. A Safire reactor costs under $20-million to
| build, and probably much less these days.
|
| The Safire reactor can keep the plasma lit and going for hours if
| not days without interruption. The Safire reactor has been around
| for over seven years now.
| aunty_helen wrote:
| There was a front page hn article this week about institutions
| keeping the reason for their creation going. ITER has always
| felt this way for me.
| pfdietz wrote:
| It's not a coincidence that the ITER project will be about
| one career long.
| dralley wrote:
| ITER has the same problem as the NIF, had they been designed a
| few years later it could be a fraction of the size and cost due
| to improvements in technology
|
| But since ITER was designed decades ago, we're stuck with a
| massive, expensive, outdated beast that's taken so long to
| build it will likely end up being lapped by other projects.
|
| At least NIF has something to show for it. ITER feels like
| building the Vasa.
| pfdietz wrote:
| ITER never made sense at all, even without improvements in
| fusion technology.
|
| And the improvements have not made tokamaks sensible. Even
| higher field magnets don't rescue the tokamak concept from
| practical irrelevancy.
| tsimionescu wrote:
| And yet ITER is the only serious attempt at fusion research
| for power generation.
|
| NIF is a nuclear weapons research program, as are all other
| (non-scam) ICF designs. Other MCF designs are either more-
| or-less legal scams (such as the MIT-derived startup
| claiming to build a working fusion power plant by the end
| of next or year or so), or woefully under funded.
| XorNot wrote:
| Yeah, it's somewhat falling for the public relations spin
| to think NIF's fusion research is meant for power
| generation. NIF's fusion research is meant to simulate
| hydrogen bomb detonation physics.
|
| The fact that they take a closer look at interesting
| power generation possibilities is a fringe-benefit:
| that's just scientists being thorough, but it's not why
| it was built. It's a bonus.
| pfdietz wrote:
| > And yet ITER is the only serious attempt at fusion
| research for power generation.
|
| I disagree, in two ways.
|
| First, ITER is itself not a serious attempt at a fusion
| research program, although there is great pretense that
| it is. There is no plausible route from ITER to a
| practical reactor, even if it achieves every one of its
| goals.
|
| Second, there are other attempts that are, IMO, much more
| promising. Helion and Zap are the two that come to mind.
| aqme28 wrote:
| I did my masters in Tokamak simulation so maybe I'm biased
| (though I am a bit of a stellarator fanboy).
|
| But I've never heard of SAFIRE. I've been on their website, and
| I can't find anything explaining what SAFIRE _is_ and
| especially nothing about why it 's so much better than a
| Tokamak. I can't find anything peer-reviewed.
|
| All their marketing materials are leaving a very bad (e.g.
| pseudoscience) taste in my mouth
| devaiops9001 wrote:
| Here is an explanation that includes raw video footage of a
| Safire type 2 chamber.
| https://vid.puffyan.us/watch?v=ZBInhPFFVog
|
| Here is what was shared, including some raw video footage, of
| what happened when a small tungsten rod was exposed to plasma
| in a Safire type 2 chamber.
| https://vid.puffyan.us/watch?v=7y46wMAHnsI
| exitheone wrote:
| Nobody has ever independently verified any of the claims Safire
| has made. It's wishful thinking at best and a scam at worst.
| devaiops9001 wrote:
| The results of the Safire type 3 reactor were verified by a
| third party lab. It's amazing that you can just post out of
| your ass like that as if you have any idea what the hell
| you're talking about.
| itishappy wrote:
| Has it been verified to actually work? A 7 year old fusion
| startup focusing energy on a documentary doesn't scream serious
| scientific research to me...
|
| https://aureon.ca/
|
| https://www.safireproject.com/
| devaiops9001 wrote:
| The Safire type 2 reactor has been working for over seven
| years now. The documentary they already made is to document
| the process of creating the Safire type 3 reactor which can
| process liquid.
|
| The results of the Safire type 3 reactor rendering
| radioactive material benign were done by a third party, which
| means independent, laboratory. They literally spell this out
| in terms that even a complete idiot can understand in the
| documentary you just pointed out. Hence why a very small part
| of their team spent some documenting what the science team
| did to make Safire type 3.
| pfdietz wrote:
| You don't need to believe in "Safire" (or any other fusion
| technology) to think tokamaks are a bad idea. They're bad all
| on their own.
| devaiops9001 wrote:
| Here is video footage of plasma in a Safire reactor causing a
| tungsten rod to rapidly decay, or at least that's what looks
| like is happening, not "vaporizing" (in scientific terms)
|
| https://vid.puffyan.us/watch?v=7y46wMAHnsI
| smaddox wrote:
| Tokamaks are 1960's technology. The future of economical fusion
| appears much more likely to be based on the field-reversed
| configuration (FRC). Helion expects to produce net positive
| energy production from a reactor designed primarily for He3
| production in the next few years: https://www.helionenergy.com/
| raverbashing wrote:
| Remains to be seen.
|
| Helion would have to ship something that gives positive net
| energy first.
|
| Sure, they have prototypes, but we don't know how well they are
| currently performing
| jszymborski wrote:
| I'm not sure the date of invention is relevant. The wheel is
| stone-age tech.
|
| What matters is how we build Tokamak's has changed. A huge
| notable difference are the magnets used for example.
| Muromec wrote:
| "Tell me comrade, how doez a tokamak reactor explode?"
| Agingcoder wrote:
| I'm not sure I'm ready to trust an ml system that will control a
| fusion power plant. The potential of the the mistake making a (
| bizarre, ml-like ) mistake seems very high to me.
___________________________________________________________________
(page generated 2024-03-02 23:01 UTC)