How Google Solved Nuclear Fusion's Big Problem
HTML-код
- Опубликовано: 1 ноя 2022
- Did you know Google's artificial intelligence company DeepMind has been working to solve one of the biggest problems in nuclear fusion?
Check out the book (affiliate link):
amzn.to/3WnA5Uj
Key source:
www.nature.com/articles/s4158... [Journal]
Future Of Fusion Energy - amzn.to/3WnA5Uj [Book]
Reinforcement Learning • MIT 6.S191 (2022): Rei... [Video]
#fusion #energy #artificialintelligence 
Наука
Thanks for all the feedback, there seems to be a few questions about the reinforcement learning, and I feel like I should have given more detail as I perhaps oversimplified it for the sake of the video - so here is a little more information!
Given experiences during the training process, the reinforcement learning system learns how to predict what reward it will receive from a given action (such as changing a magnets strength). Therefore, when it is operating in the real world it can predict what action will give it the highest reward value and choose that. Therefore, it is not so much rewarded or punished like we think of it, but it learns how to make decisions that maximise the predicted reward of its action. In this sence, a punishment is more the absence of a reward. This is a pretty tricky concept to make a short video or comment about, so I recommend watching the MIT lecture I have linked in the description if you are interested! Thanks, Ryan (Ziroth) 😀
Good you react on feedback. I still don't understand how one 'rewards' an AI, so I might look at the MIT video.
I do like short videos to the point like yours. You may want to look at your analogies; people who are interested in fusion videos and Google ai do know what a simulation is, so you don't need an analogy with the pilot. The bottle one was good, while it explained a more rate phenomenon.
it's best to not call it reward/punishment, it gives an extremely inaccurate impression. the AI is optimizing for a "reward function", it's not actually a reward like a treat for a dog. it's just a value where larger=better. the AI is simply designed to try and obtain the largest possible value, and researchers make the value awarded correlated with the real world result we want, therefore the AI achieves something aligned with our goals. that's the idea
BS because AI, deep mind, ML ... - all trained on huge stats+results, but these are 0 by now.
@@edopronk1303 simply by adding to a sum. The RL algorithm deliberately chooses a set of actions that lead to the highest sum.
I understand that. You are a really great communicator and researcher hence my desire to see what you can come up with on the dreaded "cold fusion", while there is a huge stigma attached to this technology, I suspect that it may prove to be the winner in this race.
I hate how people say a torus is donut shaped; no....a donut is torus shaped! TORUS SUPREMECY
Homer would take issue with you.
"Nuclear Fusion's Biggest Problem" is that' it's not working, so they made it work? No? Title is misleading.
biggest problem is that it's nearly impossible to remove electrons from a plasma amalgam that melts everything that comes in contact with.
@@zillibran That's surely one of the biggest problems, but THE biggest problem? I think my comment still applies. ;)
"Not working" isn't a specific problem. AI is solving the underlying issue that's preventing it from working. How is that misleading?
Well, apart from being able to source enough tritium (e.g. from a lithium blanket), which is going to be a challenge, and lifetime of materials exposed to a huge number of neutrons, what is there?
I'd say maintaining a sufficient density of plasma, at sufficient temperature, for long enough, is the big one. Without that, there is not much point!
Actually that is "only" problem
I can definitely understand why Google is interested in cheap clean energy. Their server farms take up an enormous amount of power. It would be nothing but good news for them.
FYI 20 years from now geothermal power will be the big thing. It’s going to kill off solar, wind, gas, and nuclear.
A revolution is about to kick off.
Everywhere on earth, 10km below the surface is enough heat power to produce so much power that only 0.01% would satisfy all of humanity’s power needs for two million years. The technology advances from oil & gas fracking is revolutionizing geothermal power production.
Bruh it’s good new for everyone, it would bring human to the next stage. U will not need to worry about electric bill. Invention that are too energy hungry will no longer be a problem.
I don't see how fusion will every be cheap! The capital cost of a fusion power plant I doubt will ever be feasible for anything other than prestige projects and as experimental physics.
@@prumchhangsreng979 you would still be worrying about your electricity bill because when we have more to consume we consume more. A few decades after fusion is achieved we would have things in our homes that took of orders of magnitude more power than before.
@@tjampman It produces energy energy = money. And it will do it with cheap fuel. And the military would salivate at using it in warships for cool stuff like railguns. That require immense amount of power. And many other things.
Using AI to predict anomalies and quickly adjust to confine a sustained fusion reaction? Has anyone checked in with Dr. Otto Octavius to see how his plan for this is working out for him?
Advise not connecting the AI directly to your brain, nor to giant mechanical claws.
I prefer the Barkley approach to deactivating the Argus Array fusion reactors by becoming the Enterprise's computer. Nobody dies and the crew gets a quick trip to the galactic center.
@@abj136 You joking? That's all I'm waiting for!
man... i hope you trolling , but there are still so many ppl that base theyr fear of AI on films and fantasy ...
@@unknow11712 I was joking, about becoming Octo, yes but when the tech is "safe" I'm absulutely not against nor fearful of brain-machine interfaces. I subscribe to the transhumanist philosophy.
Wait wait, I have heard this one before; is this the story about using artificial intelligence to control tritium fusion, where the lead scientist was named Dr. Octavius?
In addition to the four points you listed comma the fuel cycle being probably the hardest of what you listed, there's the issue of how you bleed off energy through the cryogenically cold the super conducting magnets to drive a steam turbine or some sort of plasma dynamic generator. There are ideas about how to do this for example Neutron capture in a sodium blanket but that's not an idea without problems.
I still think molten salt thorium reactors are a much more viable source of nearly unlimited energy for the near term.
I wondered the same thing and agree about the molten Thorium salt reactors. Practical fusion still seems decades away, while I get the impression that commercial Thorium salt reactors could be developed much faster if we put our societal / policy minds to it.
That's why I was always more interested in aneutronic fusion. Since most of the energy is release as charged particles instead of neutrons you can just use induction to harness that and convert it directly to electricity. Also has the side effect of not having to worry as much about reactor embrittlement and the fuels are more abundant and not radioactive. The main downside is that it's 10x harder to achieve temperature wise:/
@@capt4in1 I haven't seen anybody do a study on it yet but my guess is that most superconductors, particularly the high temperature one suffer a lot with prolonged radiation exposure.
@@capt4in1 I just looked it up and just this year July 2022 there was a paper published in nature about the degradation effects of neutron radiation in high temperature superconductors used for fusion. That's another reason to go a neutronic or better yet to Molten salt reactors
I'm no expert but I'm smart enough to know I know nothing, and also follow science news best I can, with some very smart friends. Last I heard Hastelloy or similar was basically perfected to contain the corrosive Thorium. Also makes me smirk when well-intended wallys of our end backfire and call Fusion a 'pipe dream'
As they have no clue the Greenwald Limit got updated, then go light up.
What I take away from this video is that via deep learning we can decrease the time we need to figure out the current limits and flaws of our models or the used method and technology, but it isn't clear wether this will ever solve anything.
So the word "solved" is totally misplaced in this video. It rather should say: This approach X could drastically increase testing speed for experimental fusion reactors.
The important bit separating reinforcement learning from other training methods is that it rewards a chain of actions rather than a single response. It is often combined with autoregressive ANN nowadays, so the network just picks one action at a time, then it is fed back its previous actions and their result(s) as input.
Still not enough though, Ai needs an actual thought mechanism, not just a action selection ability.
@@Binahx86 my guess is we'll have something like that by 2030
@@Binahx86 we already have some stuff like that, recent papers have developped on associative memory mechanisms with spiking neural networks, which tend to reproduce what actually happens in our brain (spikes of energy from electricity currents) to emulate a thought mechanism
It's still in research phase, and looking at how some convolutional neural networks and other memory based networks have only been implemented to public use recently, id guess we'll see those implemented in 5 to 10 years
Kalki Ironman type 7 and 8 after 2026 😎
kalki avatar (beast of the earth) (christ on the white horse) (son of man on clouds) is the biggest enemy of dajjal/antichrist/kali 😏
Kalki Avatar (Murtaza) 11th satguru 13th imam cousin of Moula mahdhi a.s. 12th imam (muhammad) 😎
Prophet Moula mahdhi is raja shashidhuvj (the mighty one) born less then 1200 years ago 😎
Prophet Moula Isa a.s. will kill dajjal cause dajjal is going to kill Kalki Avatar 😏
Kalki Avatar will follow orders from 2 religious king Moula mahdhi a.s. and Moula Isa a.s. 😎
Kalki Avatar going to have 2 swords and ring of moula sulaiman a.s. and staff of moula musa a.s. (iron rod) Staff of moula musa a.s. is like omintrix can transform into anything and can transform others into anything And stone in the ring of moula sulaiman a.s. is also known as kastav mani and it's more powerful than all 6 infinite stones combined 😇
Cuz Kalki is ironman batman super saiya-jin superman ben10 saitama optimus prime shaktimaan and every super heroes combined after 2026 😎
This staff will transforms into white horses with wings,weapons,iron-man,cloud etc or can do imagination into reality 😎
*Ratn sru sword of lord shiva (miri)😇
*Ratn varu (zulfakar) sword of Moula Ali (piri) 😇
miri piri 😇
Kalki Ironman after 2026 😎
Satyug (sunrise from West) 2038 😏
Sambal is hospital 😏
Gzwa e hind 2029 😎
Khalistan and Azad Kashmir after 2026 by Ironman 😎
99% Hadith u heard is not about imam Mahdi it’s about Kalki avatar (the main character) that person momin vs dajjal prove me wrong if u can 😏😏😏...
Kalki ironman after 2026 😎
Humans 0
Saiya-jin 1 to 6
Angels 7 😇
Kalki Ironman 5th matriya buddha 8th arc Angel 11th satguru 13th imam and 24th avatar after 2026 😎
Almighty God 9 😇
Humans type 2.5+ after 2026 by Kalki Ironman (christ in the white horse)(son of man on clouds) type 7 and 8 😎
Jarvis the world’s first artificial intelligence (parrot) 😎
Cuz Kalki is ironman batman super saiya-jin superman ben10 saitama optimus prime shaktimaan and every super heroes combined after 2026 😎
Jarvis world’s first artificial living being (just like vision in marvel universe)😎
Made of Quantum and sub quantum particles 😎
Power source quantum energy arc reactor (type 7) level technology 😎
Kalki Ironman going to have every kind of arc reactor like:- type 1 Nuclear fission, nuclear waste, hydrogen fusion, type 2 3 4 5 antimatter arc reactor (solid liquid gas), type 6 electro quantum arc reactor, and type 7 Quantum arc reactor, type 8 limitless sub quantum energy arc reactor without sub quantum particles 😎
Kalki Ironman going to have sun in a box million billion tons of hydrogen nuclear fusion reactor type 2 3 4 5 just like sun and stars in the palm of his hand 😎
This all going to happened by self replicating quantum nanobots knowledge energy and techniques at type 7 7 7- respectively 😎
Ironman (Tesla 2.0) going to reveal every secrets of the world specially Tesla and his Antigravity 😎😎😎
Interesting video. And nice to see your subscriber numbers grow. It'd be interesting to see a short channel update about that from time to time.
And it would also be interesting to see more about reinforcement learning. Especially how this works on a computer system, when rewards and punishments seem to make sense only when some kind of feelings are involved.
If rewards = True then
keep going
ElseIf punishments = True then
try again
End If
There are your feelings...
I have added a link in the description to MIT's great video about reinforcement learning which gives much more details! Unfortunately, it is a complex topic that is very difficult to explain accurately in a short pop-science video! I hope you find this source useful :D
@@ZirothTech Thanks, I'll pass that on 👍
@@B.Lastend a machine doesn't care about iterating as nauseam. Maybe you should watch the linked video, too.
@@human_isomer maybe I am a programmer and just wanted to explain u the video content in simple. I said nothing wrong 🤷♂️
Been saying it for years we need an ai to maintain the fusion reaction. Our computers are very fast but what the problem is that we're not doing something in the present you need to be able to predict inputs to maintain a small or proto star. That kind of of constant vigilance and insane algorithmic fore thinking is something only a machine could do.
AI might be best for both production of better tech designs, settings and as well as maintaining fusion reactions. It should perhaps be open sourced so more than just reactor's team working on it, as more thinking heads can solve problems better and faster.
Yeah and fusion scientists have been doing that for years, since 2014 at least. It’s not as simple as slapping on an ai algorithm to diagnostics since the signals coming in are discrete and not high enough resolution for good real time control
Masters degree in physics here. I enjoyed this video. You taught me some terms from the industry and explained it quickly rather than going over the absolute basics you presented something new around AI applications and what’s happening. Appreciate it
Glad you enjoyed it! I appreciate the kind words
Maybe I just missed the information, how did the AI perform with the real reactor ? I don't think you provided data on its achievements so far. So... why ''Google Solved Nuclear Fusion'' in the title? Did I miss something.
The graphs of it containing the plasma are from the real reactor, sorry that wasn't made clear
@@ZirothTech thanks for the clarification:)
Great video, please do more! I bet you could be both an awesome space sci fi channel as well something for other stuff would be cool.
4:20 "mathematically punished"....sounds like you're asking for revenge
So does this mean that commercially viable fusion is now only 30 years away?
Haha
lolz
Soooooo, that all gets us about 30-40 years from readily available industrial fusion power plants, right?
Really cool use of reinforcement learning! Thanks for another great video :)
Thanks Lois!
If I understand correctly, the Wendelstein 7X is also an attempt to shape the reactor into these idiosyncratic forms that are best suited to containing the plasma. Look it up, it’s wild looking and just ramping up
I clicked on this video thinking "Sigh. Here we go. Let's see what kind of malarkey we have here.". I was pleasantly surprise to see nothing of the sort. 🙂
Well done mate. Good video
A Tokamak means you have a natural limit to confinement time ... also there are two different temperatures --- one for the electrons and one for the nuclei --- I'd not say that this was a break through as practical fusion ( fusion with energy gain) is more complicated --- especially when scaling it down to human dimensions as giantic constructions do not have those constraints like "the monster machine" or brown dwarves/failed dwarves.
They found a sustainable and afforable Tritium source?
Breeding it in a lithium blanket or summin' like that
Breeding
I think you need to read up about the involvement of the UKAEA and The JET experiment at Culham in the UK. Jet has been going since 1992 and a huge amount of data from that has gone into the ITER build. Including the latest X-Diverter technology
If only there was a big ball of plasma in the sky that could do nuclear fusion for us :(
:chuckles: Take a guess at the efficiency rating of PV ... and it's relative cost.
I'll light a candle in support of having a big ball of plasma in the sky
@@dallassukerkin6878 :chuckles: Take a guess at the efficiency of nuclear fusion and it's relative cost XD
And if we could build a Dyson swarm around it to collect that energy, we’d be good.
However, we currently don’t have access to that.
@@Strop9911 :grins: A bit moot at the moment given that it's not a functional technology. Once it is (or IF :)), however, then humanities power problems are, if not over, then significantly resolved in terms of cost and availability. PV has it's uses but it is not efficient in either production or cost per unit.
Hmm the reason why we choose to use a tokamak is because you can change the fields very easily it’s not some that will be used for the reactors themselves that produce energy. If you thought that my bad they will use a stellerator after finding the best configuration for the magnets as it’s way better.
Plasma is super hot. But also incredibly hard to get it to that temperature. The problem of the plasma patch in the wall is not burning the wall, but cooling down the plasma.
Funny how you took a small part of a full interview I had already watched and made a whole video about it.
New subscriber.. I like what I see dude, well played.
If the plasma touches the inner walls it wouldn't burn anything. The plasma would cool down and the reaction would stop. That's also why it is very safe.
For quite some time I have imagined that someone was working on modelling and control of the plasma instabilities, but this is the first video on this very important topic I have seen
Good stuff bro. Keep going
Thanks Eric!
Even if fusion power become reality it's still dependent on the electrical grid. Solar is not. The independence solar gives is a huge advantage. In summary, yes, fusion would be great especially for the industry, but it's not really needed.
Justin ball is my professor, such a great teacher !! I appreciate the fact that he answered some of your questions, good content !
Fusions first and biggest problem is that Tritium costs $30,000/gram requiring charging a price for electricity that by far exceeds current levels.
There is a way to produce tritium inside a fusion reactor, which could make it affordable.
"how google's ai is helping to solve one of nuclear fusion"s big problems" would be a truthful title
Sounds great. What I still wanna know is, how do they expect to get usable energy out of this?
As usual like all the other power-plant technologies. We're boiling water which powers a nozzle.
The trick is to use as few input energy as nesessary to hold the fusion-reaction "going on" and harvest more energy it outputs then we used to make it start.
@@steinis6409 But how do you boil water from a magnetically-confined plasma that's at millions of degrees, inside a vacuum? It's not exactly like you can run piping for a heat exchanger through it...?
@@michaelsmith2785 One product of the fusion are neutral charged but ofcourse very hot (uncharged) neutrons. Those can leave the magnetic-vacuum due to missing charge they don't interact with the magnetic field.
Instead they will hit the mantle which is made to be hit by the neutrons. The neutrons will cool down alot when hitting the generator-wall which will heat up in exchange..
The heat from this outer wall will be transported via heat exchanger to the water which boils, powers the nozzle and so the rest is well known technique.
Building entire ecosystem around fusion reactors will be perfect in mars biospheres.
I'm glad the title isn't "biggest problem" as the absolute biggest problem with nuclear fusion is that it's always 10 years away and have been so for almost a century. Notify me, when we have a reactor actually producing power for the people.
Who says it’s ten years away?
And yet they cant write a navigation app that doesnt route me through 300 miles of small towns at night.
I thought you were going to say "How are they going to get all that Tritium." Fusion is a 100+ pipe dream; worth working on but don't make any bets on it.
“Mathematically punish” sounds like a discipline ill never get used too or understand 😞🙂😔
How do you hold the sun in your hands without getting burn? Similar question was asked over and over again in the history of Humanity. We played with fire first, with steam, with internal combustion, with nuclear fission. Now we just need build a lamp with a tiny sun in it.
"Dr Octopus it's not ready, it has a mind of its own"
A lightsabre is actually a miniature, hand held nuclear fusion reactor.
Who would've thought - the solution was just a "Google search" away?!
Why does Google solving the nuclear Fusion problem feels like the rising of SKYNET!
In a world where fusion power plants exist you would have a thermal pollution problem.
That's a fancy way to say “global warming”
"Thermal pollution" isn't a power generation problem, it would be a power consumption problem. These reactors would be very well insulated because loss of heat to the environment results in a loss of efficiency. Heat would come from all the electrical appliances connected to the plant, which turn electricity into useful energy that winds up eventually as heat.
@@Darth_Insidious thermal pollution is a power generation problem as well.
In particular, coal and gas fired power plants release a lot of energy as waste heat into the environment. Geothermal and nuclear plants release less waste heat, and solar, wind, and wave generators extract energy from the environment, so don't suffer this issue.
@@matthewparker9276 I wonder how big a problem waste heat would actually be. Seems like our concrete jungles do more to heat up the local environment than power generation.
Waste heat from power production and consumption is, at present rates, about 1/100 the scale of the heating problem caused by excess greenhouse gases. It will be a long time before we reach the point where we've decarbonised electricity, transport and industrial heat and then grown power production sufficiently to cause an actual planetary heating problem. If we get to that point, more CO2 capture would redress the balance. If we even later still get to the point where there isn't enough CO2 left in the cycle for plant life, we're probably doing all the heavy industry in orbit and further out in space, so we can reduce energy consumption needs on terra firma anyhow. A quicker fix would be to replace concrete with some kind of composite.
Could you provide confinement time before and after RL?
A really great brisk read on Fusion is “The Star Builders” by Arthur Turrell
a bit missleading to be fair.... the biggest problem of a tokamak is not the biggest problem of fusion in general.
Besides there are a number of big problems we still have. One example is the T-breeding ratio one can obtain. What point is running a tokamak at high(er) densities if you dont have the supply of fuel to begin with? Besides running it at higher densities means more energy (which is what we need) but this also increases other problems, such as plasma confinement as well as divertor heat loads. Besides, we can already run high temperature high density plasmas to begin with, so thtis is more of an increase in plasma stability rather then fixing ''the biggest problem''
The biggest problem is Tritium supply. There are only 2.5 kg available which will be consumed by experimental reactors. No more Tritium can be produced.
Another great video from Ziroth
I figured one of fusion's biggest problems would be how to get any useful energy out of the system. Like... exactly how are you extracting it and turning it into something that could say... power a city? At some point in time you need to have the actual plasma interact with something physically in order to transfer energy from one form to another, like a need to weaken or turn off the magnetic field in order to let something come through. With the operating temperatures of the plasma... that would be pretty much impossible right? Considering that we don't have any materials that could withstand direct contact for long periods of time.
Tokamaks, stellarators, etc. do not intend to extract the energy from the plasma within the confinement region. When you fuse Deuterium and Tritium nuclei, you get a helium nucleus with 1/5 of the fusion energy and a neutron with 4/5 of the fusion energy. The helium stays confined in the plasma and helps maintain the heat (saves you adding external heat to maintain the pulse) but the neutron is not confined. The idea is to capture the kinetic energy of the neutrons in a liquid lithium "blanket", outside the plasma, directly from collisions and indirectly through fission reactions the neutrons cause which, conveniently, produce more heat and more tritium. Pert of ITER's work is to figure out and prove the lithium blanket system, but other MCF projects are working on it also - CFS, Tokamak Energy, etc.
Deepmind> We will have fusion power in 30 years.
At some point they should still try to figure out how to collapse that donut hole and do field isolation without a physical central column. It's a lot of wasted volume as far as concentrating energy would be concerned. The tighter you can collapse the plasma towards a pinch-point or focus, the hotter you can get it with less energy being input - therefore more efficient.
Greetings from the COMPASS tokamak :) !
Google couldn't solve a jigsaw puzzle!
Presumably there must be many sensors to detect the shape of the magnetic field and provide feedback for the magnets. However, I have been told that such sensors, and any diagnostics with insulators, will have an extremely short lifetime in the DT neutron flux, and plasma control will go out of the window.
Here is a quote from their scientific paper:
"TCV (the reactor) is equipped with several real-time sensors and our control
policies use a subset of these sensors. In particular, we use 34 of the wire
loops that measure magnetic flux, 38 probes that measure the local
magnetic field and 19 measurements of the current in active control
coils (augmented with an explicit measure of the difference in current
between the ohmic coils). In addition to the magnetic sensors, TCV is
equipped with other sensors that are not available in real time, such
as the cameras."
I hope this helps!
Thanks for this - you’re well informed. So how will they all deal with the 14 MeV neutron damage?
@@johnh6245 isn’t neutron heating supposed to be the main goal? So a cheap modular blanket material that can be swapped out will be the goal
Maybe a plasma jet engine would be more simple to make, but who knows if you could actually get more power out than you put in.
Theres actually a lot of problems, and every single one prevents it from being viable. Im just going to mention a couple that i haven't seen in the comments. One is just the material integrity of the chamber. There arent a whole lot of materials out there that can withstand being bombarded with neutrons over an extended period of time, and the next is how to extract the energy. Its not exactly an easy thing to even think about, since any extraction method focusing on just capturimg or diverting electroms is probably going to destabalise the system in some way.
Having talked with a scientist from ITER the method of extracting energy is actually quite simple. The inner donut closest to the plasma is constantly cooled by water and the heated water drives turbines similar to a fission reactor.
@Tom's Cubes you dont need a scientist to explain how transferring of heat works. Simply watching a kettle with cold water will suffice.
@Klaus Schwab the electons, with a lot energy, smash in the wall and create heat, this heat go on water (or other liquid) to run a turbine.
i made a mistake, i think it's neutron,not electrons, who are not interact with magnetic fields because neutre.
The biggest problem with this design is simply in the middle...the center ring is poorly designed to hold PLASMA. I recommend using more natural shapes with two Pyramidion poles reflecting each other in a Torus
09/22/2020.
Clean energy.
Comprehensive precision machining for vacuum vessels, life support of sustainable nuclear fusions.
Perfect joints with controlled asymmetry for the efficient construction of hermetic vacuum chambers.
Nuclear fusion.
New automatic laser electromagnetic technology.
Tungsten. (W), austenitic steels, aluminum.
Plasma st
In StarCraft Protoss voice: "You require more tritium gas!"
150 Million degrees???? That is kind of pretty hot.
Wenn dies dann richtig abläuft, braucht man unter zusätzlichen vektoriellen Feldstärken eine richtige
Teilchen spezifische Kernresonanz (Protonen und Neutronen jeweils ) damit auch alles richtig resonant
im Kern zusammengeht.
I have a idea it's a theory so bare with me thinking outside the box. What if the inside chamber had three separate sets of magnetics that could make three separate plasma rings. Any irregularities for shifts in plasma could jump between the three rings of plasma. Obviously there would be some engineering involved. I believe a pull on one would cause a push on another. In my theory this constant push and pull between the three could be enough to have a self sustaining relationship of balance.
You would basically have stacked tokamaks. You won't be able to create three separate plasma rings in side of a single toroid.
I wish you had a discord would be very cool because I got stuff you will love to see for fusion
Dr Octavius already did this in Spiderman 2
Does sound travel through plasma? Perhaps harmonics could be used with magnetic fields to produce novel shapes.
P.s. You have to think about the SIZE,SHAPE,MATERIALS you're using there's so many things you have to think about and get just right that it just won't happen FOR A H*** of a long a** time. And when you spend the time to think about it a lot of the words that they use they talk about it essentially being a perpetual energy machine. Which from what I was always told as a kid doesn't exist
Mr picky here again ;) The word nuclear is pronounced new-clear. There is no middle syllable.
Fascinating video!
Love your animation style and diagrams!
You missed "one of the main" between "fusion's" and "big problem(and missed the S)". The shape ok, is quite a big issue. But THE main, and at the moment unsolvable one, is the fuel.
My prediction? Fusion is about twenty five years away.
Question: How is energy withdrawn from the reactor to provide energy to generators?
@@mr_clean575 I understand that but how is the heat removed from the reactor. The plasma is contained in a mag field. How is the plasma handled so that it can transfer its heat??
Wait, we're going to give an AI Nuclear Material then Punish it like a puppy?
Didn't ANYONE at google ever watch the Terminator series? ☠🤖☠
😂
Not literally punish.
I feel bad for the AI being mathematically punished :(
Don't. It will one day kill us all.
I'd like to remind everyone that there are two energy in/out numbers. The reactor and the entire facility. We are a long, long way from fusion power plants even at a 20x payout from the reactor due to costs running the facility.
It's important research not a golden bullet.
Super cool!!! 😁
I think your reactor is still far away from the target. I studied these things in 1973 and then there were similar promises but still nothing has come out for practical energy production. There are simply too many severe problems unsolved.
I am sure the fusion will really come out at some point in the future. We are not there yet.
Very interesting video. Though I don't get why the word "nuclear" is so hard to say. Basically nu + clear.
There's no ai in the world that could ever make a broken fusion design work. The problem with fusion comes from the lack of scale. And no computer program can overcome that.
Make a video on quantum noise power device..
(This is a comment section, and I’m not citing any sources, so I hope you understand this is my personal opinion.)
I find it weird that people keep calling it reward or punishment. It is not punishment. There is no pain, no negative consequences for the algorithm. It is only picking the one that we give a higher value as a result.
It is just like saying pick the closest to this point. Nothing else, no rewarding, no punishment, there is no sense of self, or gratification or enjoyment. Just directions.
I also really don't have such SUPER-CALCULATOR, as sophisticated as Your AI .
With the cost of home based solar pannels and batteries dropping below the cost of building power lines, how is fusion going to compete?
If this worked, it already would have. It's obvious the human intelligence involved, isn't high enough.
By the sounds of this it should be "about another 20 years" and fusion will be ours.
Well, good to know there is some use out of Google still. I was starting to think this company will only focus on helping setting up a surveillance state, and developing intrusive, unavoidable advertisiment with data mining.
Google is primarily an artificial intelligence company. Always has been. Advertising is just one use of many for their models. Their focus isn't on primarily selling user data, so unlike Facebook Google will be fine even with Apple becoming increasingly closed off with data collection on their systems.
Yeah ...the Long Now foundation is pretty cool too...hope I live long enough to see fusion become a viable power source...
@@Darth_Insidious google made 80-85% of their revenue through advertisement. Google is pretty clearly primarily an advertisement company with a gigantic focus on indirectly selling user data.
@@Darth_Insidious Google operates the exact same way as facebook, they utilize user data to deliver targeted ads which in turn generates more through advertising. They're going to take a revenue hit same as facebook.
Bold of you to assume a tree that only produces evil fruits will spontaneously begin producing good fruit as soon as they have the technology that could be used to effectively enslave the rest of the free world by energy monopoly...
The fusion reactor looks alot like the big arc reactor in iron man 1.
Could the process of fusion nbe further used for synthesis of lithium, or any such element in short supply?
It will actually use lithium which produces the tritium fuel when hit by a neutron. However because the fuel is so energy dense it would arguably be a more efficient use of lithium than other uses. It might be efficient enough that you could cost effectively mine trace lithium out of seawater
All the fusion fuels primarily produce helium but because they are so fuel efficient they wouldn't make very much. Above helium it would be very expensive to put in the energy to produce heavier elements as fusion becomes much harder when you go past helium
Not going to lie I thought for a second that you showed us a recipe book on doughnuts then😂
OK, so you've got the plasma contained by super-cooled magnets that have to be reasonably close to the plasma, but to run steam turbines you need to extract a huge amount of heat to boil water to drive the turbines. So how are the magnets to be kept cool, whilst all this heat is coming of the fusing plasma ? Bear in mind this heat extraction has to be continuous for weeks, even months at a time.
Fission reactors use the reactor coolant, British gas-cooled reactors with carbon dioxide gas and American and others by water. Both pass this coolant through heat exchangers that are the actual boilers that produce the steam.
So what is the equivalent process in the tokamac. I have yet to see anything that proposes how this heat extraction on the scale needed, is done.
This is a great question Fraser, and I think it will be the topic of my next fusion video! There is a good little section on the 'Fusion Power' Wikipedia page called 'Energy Capture' that has some information. The key idea is the high energy neutrons that are shot off from the fusion reaction fly to the edge of the reactor wall, as they are not affected by the magnetic fields. These neutrons then collide with a protective blanket and heat it up, this blanket also stops damage to the magnets. The hot blanket can then be cooled with a fluid, extracting heat for a steam turbine.
The magnets are cooked with liquid helium. They’re in the cryostat, away from the blankets, so it’s pretty easy.
Need to cool the area outside of mag field this can bee done by cooling and keep wall somehow removing heat or another fealty out side of plasma which can distort heat and cool it down a program made to control heat and move it or to keep wall cooled whith coolent of som sort if wall can be destorted to absorb and remove heat rapidly
well done
"This is similar to pilots training on a flight sim: its cheaper to run."
...Dude, both are so that way the trainee doesn't crash and burn. As expensive as it is to run a fusion reactor, it is hundreds of thousands times cheaper than having to build a new reactor. And, ya know, it keeps the facility from turning into a small hydrogen bomb if the AI goes tits-up in the early training stages. You don't have to frame the problem in terms of cost vs runtime.
Also, most fusion reactors physically cant run more than a single second not because of cost, but because they don't want to turn the damn thing into a slag heap from the temperature. Because current reactors have a bad energy confinement, with some of those millions of degrees of temp leaking into the reactor body.
How do you "mathematically punish" a machine learning system? How does that work exactly?
It could just be -1 for some arbitraty score
Plasma stability is certainly a limit to the ratio of the plasma to the magnetic energy hence the ability of insulating the reacting fuel in a given volume.
Is this what limits at present the potential of magnetic confinement fusion? Absolutely not, what pushes the cost of the fusion kWh beyond any reasonable competitiveness to other form of energy, even fission energy, are others: tritium breeding, first wall interaction, neutron induced brittleness, operational complexity, capital cost, armaments proliferation and many others. None of these issues seems to be a good candidate for AI
Thanks for your comment! I also think there are many challenges that AI won't help with much, but I think it will be a useful tool for solving certain problems, such as plasma control!
@@ZirothTech plasma control has been a computerized task in most magnetic confinement experiments for the last 50 years, most plasma position and shape need active feedbacks which are mostly done digitally, in later years I have never seen any original AI application in this field to speak of. I think that this research is so primitive that before planning for a new experiment many virtual ones should be investigated down to a practical power plant level. The title of this video has nothing to do with reality.
Containing the plasma is one problem. Containing high energy neutrons shredding your containment is an entirely different issue. There is another fusion chain that might be used to reduce the high energy neutron bombardment, but it involves lots of something when we don't have - lots of helium 3.
Neutrons are intended to be captured to produce T from Li.
.. which involves building a moon base to obtain
So future of energy is super AI controlled giant magnet torus nursing an artificial star fueled by moon stones
@@alpers.2123 Close. The future of energy is a large monopolistic utility company owning a single expensive point of electrical supply.
Sort of like what we have now, but with lasers and magnets.
@@47f0 reminds me a couple of sci-fi movies