Plans that require an exaFLOP AI supercomputer first are broken nuclear fusion plans

Researchers at the U.S. Department of Energy’s (DOE) Princeton Plasma Physics Laboratory (PPPL) and Princeton University will use the exaFLOP Aurora supercomputer starting in 2021 to help solve the problems and complexities associated with the ITER tokomak project.

Regular Nextbigfuture reader Goatguy notes:

we have yet to invent a general AI that can even pass the Turing Test, let alone figure out how to make FUSION, upon which the top 100,000 minds in Physics from around the world have pitted themselves for the last 50+ years, to no avail.

Think about that!!! 5,000,000 (or more) seriously talented “GENIUS” level intellect-years of investment and we haven’t cracked “cheap, ready, potent, ubiquitous and limitless” fusion power. Seriously: its not tho’ we’ven’t tried. Five million genius-years is an awful lot of horsepower.

By comparison, the entire APOLLO space program had an investment of less than 700,000 GENIUS years of intellectual investment. Some would say, critiquing the armies of pocket-protector bearing munchkins that endlessly filled out gub’mint forms, reports, and documentation records, that we had less than 100,000 genius-years of true investment. (My Uncle … the “really damned smart one” figured deep in NASA and the Moonshot program. He had a LOT to say regarding the average intellect, the average proficiency of mushroom employed by the Big N.)

But FIVE MILLION genius-intellect years of progress, slow that it very well has been, is somehow going to be upended and reinvented .. correctly .. by a bunch of AI ‘bots that HAVE YET TO BE CODED?

Nextbigfuture notes that if your 30+ year technological development project hinges upon yet to be developing new more powerful superconducting magnets and an exaFLOP supercomputer that were NOT part of the original project plan then the plan was broken decades earlier. There was promise in the original Tokomak research but the research has made the project more and more terrible.

The expected cost of ITER has risen from US$5 billion to US$20 billion, and the timeline for operation at full power was moved from the original estimate of 2016 to 2027.

The 14 MeV neutrons produced by the fusion reactions will damage the materials from which the reactor is built. Research is in progress to determine whether and how reactor walls can be designed to last long enough to make a commercial power station economically viable in the presence of the intense neutron bombardment.

There are dozens of other options for commercial nuclear fusion. We need to retreat to consider whether other technological paths to nuclear fusion or whether advanced molten salt fission are the better answers to breakthrough energy.

The planned start of deuterium–tritium operation is now 2035.

ITER will not generate commercial power.
It will have have the follow DEMO pilot reactors project to prove out fusion at commercial scale then those will be followed by commercial nuclear fusion tokomaks.

Those reactors are currently projected to be football or baseball stadium sized machines.

65 thoughts on “Plans that require an exaFLOP AI supercomputer first are broken nuclear fusion plans”

  1. Almost from the start it was clear that fission was the better approach. Fusion is difficult to achieve under the best of circumstances, building equipment that will survive decades while producing it is going to be crazy difficult. While fission is so easy it almost happens on its own. The obstacles to fusion are unsolved physics, the obstacles to fission are just material science. And fusion faces those problems, too, once the physics is solved. Finally, we’ve known for half a century how to successfully cause fusion with excellent breakeven to occur, in bombs. And for almost all of that time have known that power plants could be based off those bombs. And yet we have expended huge resources on far more dubious approaches to fusion. Why? It’s the best which was designated to be the enemy of the good enough fission, that’s my conclusion. Just imagine how well developed fission power would be today, if all the work that’s been expended on fusion had gone instead to improving fission. What a waste!

  2. The expected cost of ITER has risen from US$5 billion to US$20 billion, and the timeline for operation at full power was moved from the original estimate of 2016 to 2027.” Duh! ITER is to science what the High Speed Train to Nowhwere is to California what ULA/SLS is to NASA what SOEs are to the Chinese Economy: Nothing but pork-barrel, feather-bedding, vote buying Big Science Grant Whoring 101, kiddies. But then again, there are Kool-Aid drinkers on here who will scream otherwise. 5..4..3..2.. “ITER will not generate commercial power. It will have have the follow DEMO pilot reactors project to prove out fusion at commercial scale then those will be followed by commercial nuclear fusion tokomaks. Those reactors are currently projected to be football or baseball stadium sized machines.”” Duh! Re-read what I said in the above…

  3. AI is all about reducing a system to the bare minimum number of features necessary to create a reasonably accurate model of the the system. So can nuclear fusion in a tokomak be accurately modeled with anything other than fusion in a tokomak ? I don’t know. But just saying you are going to do it with AI and some hand waving is silly. We don’t yet have a general purpose AI because no one yet has developed a universal learning node where you just add more nodes to scale with what you are modeling. We do have artificial neural networks that can be made to work after a lot of head scratching and hair pulling out. But as of now you can’t explain why one approach worked and another did not.

  4. Almost from the start it was clear that fission was the better approach. Fusion is difficult to achieve under the best of circumstances building equipment that will survive decades while producing it is going to be crazy difficult.While fission is so easy it almost happens on its own.The obstacles to fusion are unsolved physics the obstacles to fission are just material science. And fusion faces those problems too once the physics is solved.Finally we’ve known for half a century how to successfully cause fusion with excellent breakeven to occur in bombs. And for almost all of that time have known that power plants could be based off those bombs. And yet we have expended huge resources on far more dubious approaches to fusion. Why?It’s the best which was designated to be the enemy of the good enough fission that’s my conclusion.Just imagine how well developed fission power would be today if all the work that’s been expended on fusion had gone instead to improving fission. What a waste!

  5. The expected cost of ITER has risen from US$5 billion to US$20 billion” and the timeline for operation at full power was moved from the original estimate of 2016 to 2027.””Duh! ITER is to science what the High Speed Train to Nowhwere is to California what ULA/SLS is to NASA what SOEs are to the Chinese Economy: Nothing but pork-barrel”” feather-bedding vote buying Big Science Grant Whoring 101 kiddies.But then again”” there are Kool-Aid drinkers on here who will scream otherwise. 5..4..3..2..””””ITER will not generate commercial power.It will have have the follow DEMO pilot reactors project to prove out fusion at commercial scale then those will be followed by commercial nuclear fusion tokomaks.Those reactors are currently projected to be football or baseball stadium sized machines.””””””””Duh! Re-read what I said in the above…”””

  6. Some one may say that the Turng test has little to do with cracking nuclear fusione problems, which may be heavily reliant on raw calculations. Same as saying that not being able to crack the Tuin g means computers are not good in calculations Goat Guy should not be taken as the Bible, especially after he said that planes fly as they “push air down” or that it would take some $3M to use the Falcon BFR to fly one person from one side of the globe to the other

  7. AI is all about reducing a system to the bare minimum number of features necessary to create a reasonably accurate model of the the system. So can nuclear fusion in a tokomak be accurately modeled with anything other than fusion in a tokomak ? I don’t know. But just saying you are going to do it with AI and some hand waving is silly. We don’t yet have a general purpose AI because no one yet has developed a universal learning node, where you just add more nodes to scale with what you are modeling. We do have artificial neural networks that can be made to work after a lot of head scratching and hair pulling out. But as of now you can’t explain why one approach worked and another did not.

  8. Some one may say that the Turng test has little to do with cracking nuclear fusione problems which may be heavily reliant on raw calculations. Same as saying that not being able to crack the Tuin g means computers are not good in calculations Goat Guy should not be taken as the Bible especially after he said that planes fly as they push air down”” or that it would take some $3M to use the Falcon BFR to fly one person from one side of the globe to the other”””

  9. At first I thought the “AI” part was just sloppy reporting. After reading the original link, it turns out they want to use deep learning to predict plasma disruptions. That may indeed be doable, and in no way requires passing the Turing test or anything like that.

  10. We re not that far away from a universal quantum computer that can pretty much simulate anything, in a reasonable time frame. When that time comes, and it will, the quantum computer will probably show us several ways of creating fusion with the technology that is currently available. Perhaps there is even a way of creating cold fusion, that is unlike anything we have ever tried.

  11. At first I thought the AI”” part was just sloppy reporting. After reading the original link”” it turns out they want to use deep learning to predict plasma disruptions. That may indeed be doable”” and in no way requires passing the Turing test or anything like that.”””

  12. We re not that far away from a universal quantum computer that can pretty much simulate anything in a reasonable time frame. When that time comes and it will the quantum computer will probably show us several ways of creating fusion with the technology that is currently available. Perhaps there is even a way of creating cold fusion that is unlike anything we have ever tried.

  13. because the whole experiment becomes modestly radioactive after, and we can’t have our selfies taken sitting in the main chamber.” To be fair, they’re far enough from having anything working that this would likely not tell them anything, while rendering the equipment harder to work on.

  14. the price you have quoted is for the full thing. If you run BFR rocket many times you will not have to change the rocket EVERYTIME That is, reusability, the whole concept behind the Space X thing You put the money for the fuel and some maintenance and this is it. you know Less than $1M per flight , 300 people there means $3K per flight

  15. ⊕1 from me, too. Things to think about… [1] WILL IT EVER GET SIMPLE? That’s kind of a big point. Yes, there are miles of tubes and pipes, precision fuel rods by the thousands, moderators, reflectors, metrology and more to get FISSION to happen. Yet as you so eloquently reduce it, “it almost happens on its own”. In fact it does! One need only cobble together a variable density rack of rods in a neutron moderating medium (water), and have it nominally “just below” critical density (“OFF”). Then, on cue, add in the 0.1% just-a-bit-more-stuff to the rack, and watch it suddenly start glowing bright blue from the chain reaction. Added-in slowly enough, the almost-instant heating of the rods causes their expansion, which then drives the chain reaction away from super-criticality. It can be operated fairly safely. By definitely non-suicidal competent operators. Will FUSION ever get that simple? [2] WILL FUSION EVER FIND A COST-EFFECTIVE SCALE? … Seriously, this is to me the second biggest of its problems. Early work — in the 1960s and especially 1970s — showed that fusion’s “problems” with plasma instabilities rise as plasma temperature increases, as plasma density increases and as retention time increases. With no small irony, for fusion to become “a thing”, E (temp) needs to rise, D (density) needs to rise, and T (time) needs to be longer. Well. Bad confluence of angry gods. But the fine armies of researchers in turn found that making fusion setups BIGGER had positive good effect on solving all 3 parameter scaling issues. Somehow, as fusion reactors of the Tokamak design were getting larger, somehow the plasmas were become more stable, hotter, offering longer retention times and so on. All the right direction! Only problem is… the reactors are superscaling money-wise. EXPENSIVE. [3] MIRED IN FEAR OF RADIATION ISSUES — to mis this is most disappointing. The researchers have become such apparatchiks-to-conservatıve-incremental-science that they

  16. Perhaps a less tedious way to write that link: youtube /watch?v=dfY5ZQDzC5s (just strip the prefix, and replace the dot-com with a space). (Adding some extra text so that vuukle won’t count this as spam. This really is a circus of a commenting system, and we’re the ones jumping through hoops. It still wants more text. Ignore the brown fox jumping through hoops over the last dog.)

  17. If wishes were dimes, I’d be a rich man in my prime. Nice summary of the Much Ballyhooed future. Let’s see if it comes to pass. I think NOISE is going to scuttle the Universal Quantum Computer just as nonlinear instabilities have scuttled all hot-fusion efforts to date. Same issues, actually. GoatGuy

  18. I agree: it was the natural reduction to simplify everything as familiar click-bait … which seems to be the state of affairs these days across all media. I summarized it best in reply to WarrenTheApe. Go see… GoatGuy

  19. Couldn’t have said it better. ITER has become what virtually any relatively conservatıve Big Science (somewhat skeptical) analyst feared. A giant magnet (ahem) for near-limitless international Big Government Money spending. Spending on most anything that substantiates continued spending, stretching already picayune “lofty” science goals, and keeping “the patient” on life support for as long as it garners funding. “So, Joe… we have this big computer” “Yah, Bob? We have this huge reactor mock up” “Well, we think it is fast-and-smart enough to help you solve your stability problems” “Yah, Bob? We sure could use one of those ourselves. Could we buy some time on yours?” “Sure, Joe. Just put together a preliminary proof-of-requirements report, form J–1362–2017°F …” “Great Bob! Be on your desk within a month. Wanna do a press release?” “Awesome, Joe. Put that friendlier-face on the much maligned DoD” “Right, Bob. I’ll get Terry and Sandy in marketing to give your team a call.” “Good seeing you, Joe.” Something like that. THIS IS NOT TO SAY that using fast realtime computing methods mightn’t tease out of the prodigious data some previously unknown body of triggering evidence giving rise to plasma instabilities. And perhaps those newly discovered plasma burbles might later be something which electric and magnetic field modulation could counteract before they “upset the applecart”. But somehow … this old goat thinks that requiring exaFLOPS of computing power, taking hundreds of megawatts of energy to keep powered up, is not a terribly “commercial” reality. And no, I’d not be inclined to fall down the Quantum Computing is Around The Corner rabbit hole either. Just saying. You hit the nail on the head. ITER is a “Big Government Festooning Junket” if there ever was one. Din’t start out that way. GoatGuy

  20. Since VUUKLE chops out links, you’ll have to edit out the spaces to play it: https :// www . Youtube . com / watch ? v = dfY5ZQDzC5s … 0:22 → 0:50 … 1:13 → 1:45 … 2:05 → 2:30 … 2:37 → 2:43 … 3:11 → 3:27 … 5:37 → 5:54 SpaceX $25,000,000 to $75,000,000 for one intercontinental ballistic SpaceX people-mover. $125 M – $75 M / ( 10 to 100 people ) → $1.2 to $7.5 million per person Geometric mean… √( 1.2 × 7.5 ) → 3 million. Regular “average” would be higher ($4.35 mil) You’re acting the role of the Rubber Duck, sir quisling. Quack, quack … quack … quack. GoatGuy

  21. There is a tendency to think that AI can solve everything certainly strong AI. We even see evidence of this in experiments where children have demonstrated more faith what they are told by robots than by humans. Personally I think it is likely that AI (or strong AI as they are two completely different things) is more likely to allow a complete takeover of the world economy before it will figure out our biggest problems in physics for us.I believe too that the stultifying effects of an entrenched bureaucracy on creative and scientific research and development processes can not be underestimated. Some people lament the fact it’s been over a century since we had any really huge breakthroughs on the level of general relativity and quantum physics. But I think it is highly likely we’ve given ourselves some huge disadvantages in trying to do so regardless of how well intentioned they were.

  22. because the whole experiment becomes modestly radioactive after” and we can’t have our selfies taken sitting in the main chamber.””To be fair”” they’re far enough from having anything working that this would likely not tell them anything”” while rendering the equipment harder to work on.”””

  23. the price you have quoted is for the full thing. If you run BFR rocket many times you will not have to change the rocket EVERYTIME That is reusability the whole concept behind the Space X thing You put the money for the fuel and some maintenance and this is it. you know Less than $1M per flight 300 people there means $3K per flight

  24. ⊕1 from me too. Things to think about…[1] WILL IT EVER GET SIMPLE? That’s kind of a big point. Yes there are miles of tubes and pipes precision fuel rods by the thousands moderators reflectors metrology and more to get FISSION to happen. Yet as you so eloquently reduce it it almost happens on its own””. In fact it does! One need only cobble together a variable density rack of rods in a neutron moderating medium (water)”””” and have it nominally “”””just below”””” critical density (“”””OFF””””). Then”” on cue add in the 0.1{22800fc54956079738b58e74e4dcd846757aa319aad70fcf90c97a58f3119a12} just-a-bit-more-stuff to the rack and watch it suddenly start glowing bright blue from the chain reaction. Added-in slowly enough the almost-instant heating of the rods causes their expansion which then drives the chain reaction away from super-criticality. It can be operated fairly safely. By definitely non-suicidal competent operators. Will FUSION ever get that simple? [2] WILL FUSION EVER FIND A COST-EFFECTIVE SCALE? … Seriously”” this is to me the second biggest of its problems. Early work — in the 1960s and especially 1970s — showed that fusion’s “”””problems”””” with plasma instabilities rise as plasma temperature increases”” as plasma density increases and as retention time increases. With no small irony”” for fusion to become “”””a thing”””””” E (temp) needs to rise D (density) needs to rise and T (time) needs to be longer. Well. Bad confluence of angry gods. But the fine armies of researchers in turn found that making fusion setups BIGGER had positive good effect on solving all 3 parameter scaling issues. Somehow as fusion reactors of the Tokamak design were getting larger somehow the plasmas were become more stable hotter offering longer retention times and so on. All the right direction!Only problem is… the reactors are superscaling money-wise. EXPENSIVE.[3] MIRED IN FEAR OF RADIATION ISSUES — to mis this is most disappointing. The resea”

  25. Perhaps a less tedious way to write that link: youtube /watch?v=dfY5ZQDzC5s (just strip the prefix and replace the dot-com with a space).(Adding some extra text so that vuukle won’t count this as spam. This really is a circus of a commenting system and we’re the ones jumping through hoops. It still wants more text. Ignore the brown fox jumping through hoops over the last dog.)

  26. If wishes were dimes I’d be a rich man in my prime. Nice summary of the Much Ballyhooed future. Let’s see if it comes to pass. I think NOISE is going to scuttle the Universal Quantum Computer just as nonlinear instabilities have scuttled all hot-fusion efforts to date. Same issues actually. GoatGuy

  27. I agree: it was the natural reduction to simplify everything as familiar click-bait … which seems to be the state of affairs these days across all media. I summarized it best in reply to WarrenTheApe. Go see…GoatGuy

  28. Couldn’t have said it better. ITER has become what virtually any relatively conservatıve Big Science (somewhat skeptical) analyst feared. A giant magnet (ahem) for near-limitless international Big Government Money spending. Spending on most anything that substantiates continued spending stretching already picayune lofty”” science goals”””” and keeping “”””the patient”””” on life support for as long as it garners funding. “”””So”””” Joe… we have this big computer””””””””Yah”””” Bob? We have this huge reactor mock up””””””””Well”””” we think it is fast-and-smart enough to help you solve your stability problems””””””””Yah”””” Bob? We sure could use one of those ourselves. Could we buy some time on yours?””””””””Sure”” Joe. Just put together a preliminary proof-of-requirements report”” form J–1362–2017°F …””””””””Great Bob! Be on your desk within a month. Wanna do a press release?””””””””Awesome”””” Joe. Put that friendlier-face on the much maligned DoD””””””””Right”””” Bob. I’ll get Terry and Sandy in marketing to give your team a call.””””””””Good seeing you”””” Joe.””””Something like that. THIS IS NOT TO SAY that using fast realtime computing methods mightn’t tease out of the prodigious data some previously unknown body of triggering evidence giving rise to plasma instabilities. And perhaps those newly discovered plasma burbles might later be something which electric and magnetic field modulation could counteract before they “”””upset the applecart””””. But somehow … this old goat thinks that requiring exaFLOPS of computing power”” taking hundreds of megawatts of energy to keep powered up”” is not a terribly “”””commercial”””” reality. And no”””” I’d not be inclined to fall down the Quantum Computing is Around The Corner rabbit hole either. Just saying.You hit the nail on the head.ITER is a “”””Big Government Festooning Junket”””” if there ever was one.Din’t start out that way. GoatGuy”””””””

  29. Since VUUKLE chops out links you’ll have to edit out the spaces to play it:https :// www . Youtube . com / watch ? v = dfY5ZQDzC5s… 0:22 → 0:50… 1:13 → 1:45… 2:05 → 2:30… 2:37 → 2:43… 3:11 → 3:27… 5:37 → 5:54SpaceX$25000000 to $75000000 for one intercontinental ballistic SpaceX people-mover.$125 M – $75 M / ( 10 to 100 people ) → $1.2 to $7.5 million per personGeometric mean… √( 1.2 × 7.5 ) → 3 million. Regular average”” would be higher ($4.35 mil)You’re acting the role of the Rubber Duck”” sir quisling. Quack”” quack … quack … quack.GoatGuy”””””””

  30. Real, strong AI would tell you that Gen IV nuclear fission will be viable and big ITER/DEMO Tokamaks will never be viable. They can never produce cost effective power.

  31. There is a tendency to think that AI can solve everything, certainly strong AI. We even see evidence of this in experiments where children have demonstrated more faith what they are told by robots than by humans. Personally, I think it is likely that AI (or strong AI as they are two completely different things) is more likely to allow a complete takeover of the world economy before it will figure out our biggest problems in physics for us. I believe, too, that the stultifying effects of an entrenched bureaucracy on creative and scientific research and development processes can not be underestimated. Some people lament the fact it’s been over a century since we had any really huge breakthroughs on the level of general relativity and quantum physics. But I think it is highly likely we’ve given ourselves some huge disadvantages in trying to do so, regardless of how well intentioned they were.

  32. Real strong AI would tell you that Gen IV nuclear fission will be viable and big ITER/DEMO Tokamaks will never be viable. They can never produce cost effective power.

  33. The micro-instabilities associated with a tokamak has never been adequately described or measured. All these years later, the bureaucrat’s solution is to ask for more and more money to build bigger and bigger systems. We have the military-industrial complex. We also have the scientific-congressional complex.

  34. The micro-instabilities associated with a tokamak has never been adequately described or measured. All these years later the bureaucrat’s solution is to ask for more and more money to build bigger and bigger systems. We have the military-industrial complex. We also have the scientific-congressional complex.

  35. The micro-instabilities associated with a tokamak has never been adequately described or measured. All these years later, the bureaucrat’s solution is to ask for more and more money to build bigger and bigger systems. We have the military-industrial complex. We also have the scientific-congressional complex.

  36. The micro-instabilities associated with a tokamak has never been adequately described or measured. All these years later the bureaucrat’s solution is to ask for more and more money to build bigger and bigger systems. We have the military-industrial complex. We also have the scientific-congressional complex.

  37. The micro-instabilities associated with a tokamak has never been adequately described or measured. All these years later, the bureaucrat’s solution is to ask for more and more money to build bigger and bigger systems. We have the military-industrial complex. We also have the scientific-congressional complex.

  38. Just an idea: we know fusion is possible, we’re not there ( = economically over unit) yet because of (enormous oversimplification) a problem with materials (magnet not strong enough, walls not durable enough, etc…). NOW, IMAGINE THAT – not with an ExaFlop computer with shiny brand-new AI – but with a quantum computer nonetheless, one that can simulate tens of thousands of electron orbitals that connect always the same 118 elements made out of always the same protons and neutrons, we can manage to expand the amount of know substances from the current 10 million to, say, 100 million. There are some revolution-enabling materials in there, not just for energy, but also fro transportation (flying cars? Space trucks? You got’em!), healthcare, [military?], nutrition, etc…. As far as I can remember, with a fairly simple quantum processor last year IBM was able to simulate just a bunch of orbitals, so just simple-ass molecules. But qubit capacity doubles every x years or so, so aren’t we just a few years away from discovering some spectacularly working metal alloys, superconductors, new crystals for lasers, super magnets, etc… FOR REAL THIS TIME? With the right material, fusion is kids play, ain’t it? Google: Qubit-Projections-verrsus-Algorithm-Requirements-September-12-2016.jpg On IBM’s website: /blogs/research/2017/09/quantum-molecule

  39. Just an idea: we know fusion is possible we’re not there ( = economically over unit) yet because of (enormous oversimplification) a problem with materials (magnet not strong enough walls not durable enough etc…). NOW IMAGINE THAT – not with an ExaFlop computer with shiny brand-new AI – but with a quantum computer nonetheless one that can simulate tens of thousands of electron orbitals that connect always the same 118 elements made out of always the same protons and neutrons we can manage to expand the amount of know substances from the current 10 million to say 100 million. There are some revolution-enabling materials in there not just for energy but also fro transportation (flying cars? Space trucks? You got’em!) healthcare [military?] nutrition etc….As far as I can remember with a fairly simple quantum processor last year IBM was able to simulate just a bunch of orbitals so just simple-ass molecules. But qubit capacity doubles every x years or so so aren’t we just a few years away from discovering some spectacularly working metal alloys superconductors new crystals for lasers super magnets etc… FOR REAL THIS TIME? With the right material fusion is kids play ain’t it?Google: Qubit-Projections-verrsus-Algorithm-Requirements-September-12-2016.jpgOn IBM’s website: /blogs/research/2017/09/quantum-molecule

  40. Just an idea: we know fusion is possible, we’re not there ( = economically over unit) yet because of (enormous oversimplification) a problem with materials (magnet not strong enough, walls not durable enough, etc…).
    NOW, IMAGINE THAT – not with an ExaFlop computer with shiny brand-new AI – but with a quantum computer nonetheless, one that can simulate tens of thousands of electron orbitals that connect always the same 118 elements made out of always the same protons and neutrons, we can manage to expand the amount of know substances from the current 10 million to, say, 100 million. There are some revolution-enabling materials in there, not just for energy, but also fro transportation (flying cars? Space trucks? You got’em!), healthcare, [military?], nutrition, etc….
    As far as I can remember, with a fairly simple quantum processor last year IBM was able to simulate just a bunch of orbitals, so just simple-ass molecules. But qubit capacity doubles every x years or so, so aren’t we just a few years away from discovering some spectacularly working metal alloys, superconductors, new crystals for lasers, super magnets, etc… FOR REAL THIS TIME? With the right material, fusion is kids play, ain’t it?

    Google: Qubit-Projections-verrsus-Algorithm-Requirements-September-12-2016.jpg

    On IBM’s website: /blogs/research/2017/09/quantum-molecule

  41. Real, strong AI would tell you that Gen IV nuclear fission will be viable and big ITER/DEMO Tokamaks will never be viable. They can never produce cost effective power.

  42. There is a tendency to think that AI can solve everything, certainly strong AI. We even see evidence of this in experiments where children have demonstrated more faith what they are told by robots than by humans. Personally, I think it is likely that AI (or strong AI as they are two completely different things) is more likely to allow a complete takeover of the world economy before it will figure out our biggest problems in physics for us.

    I believe, too, that the stultifying effects of an entrenched bureaucracy on creative and scientific research and development processes can not be underestimated. Some people lament the fact it’s been over a century since we had any really huge breakthroughs on the level of general relativity and quantum physics. But I think it is highly likely we’ve given ourselves some huge disadvantages in trying to do so, regardless of how well intentioned they were.

  43. “because the whole experiment becomes modestly radioactive after, and we can’t have our selfies taken sitting in the main chamber.”

    To be fair, they’re far enough from having anything working that this would likely not tell them anything, while rendering the equipment harder to work on.

  44. the price you have quoted is for the full thing.
    If you run BFR rocket many times you will not have to change the rocket EVERYTIME
    That is, reusability, the whole concept behind the Space X thing
    You put the money for the fuel and some maintenance and this is it. you know
    Less than $1M per flight , 300 people there means $3K per flight

  45. ⊕1 from me, too. Things to think about…

    [1] WILL IT EVER GET SIMPLE? That’s kind of a big point. Yes, there are miles of tubes and pipes, precision fuel rods by the thousands, moderators, reflectors, metrology and more to get FISSION to happen. Yet as you so eloquently reduce it, “it almost happens on its own”. In fact it does! One need only cobble together a variable density rack of rods in a neutron moderating medium (water), and have it nominally “just below” critical density (“OFF”). Then, on cue, add in the 0.1% just-a-bit-more-stuff to the rack, and watch it suddenly start glowing bright blue from the chain reaction. Added-in slowly enough, the almost-instant heating of the rods causes their expansion, which then drives the chain reaction away from super-criticality. It can be operated fairly safely. By definitely non-suicidal competent operators.

    Will FUSION ever get that simple?

    [2] WILL FUSION EVER FIND A COST-EFFECTIVE SCALE? … Seriously, this is to me the second biggest of its problems. Early work — in the 1960s and especially 1970s — showed that fusion’s “problems” with plasma instabilities rise as plasma temperature increases, as plasma density increases and as retention time increases. With no small irony, for fusion to become “a thing”, E (temp) needs to rise, D (density) needs to rise, and T (time) needs to be longer. Well. Bad confluence of angry gods.

    But the fine armies of researchers in turn found that making fusion setups BIGGER had positive good effect on solving all 3 parameter scaling issues. Somehow, as fusion reactors of the Tokamak design were getting larger, somehow the plasmas were become more stable, hotter, offering longer retention times and so on. All the right direction!

    Only problem is… the reactors are superscaling money-wise. EXPENSIVE.

    [3] MIRED IN FEAR OF RADIATION ISSUES — to mis this is most disappointing. The researchers have become such apparatchiks-to-conservatıve-incremental-science that they don’t want to try to actually fuse anything. Just fill the tanks with hydrogen and helium, array scores of researchers, dozens of coöpting experiments around The Big Donut, and give her a blast every few weeks. Then go off, “analyze” results, and write PhD papers.

    Wouldn’t dare put in tritium and deuterium, which together stand a randy chance of actually fusing. And producing showers of neutrons. And a large pulse of thermal energy. No, no, no. Not now… because the whole experiment becomes modestly radioactive after, and we can’t have our selfies taken sitting in the main chamber.

    ———

    Those are what I see as the problems with Fusion Science today.
    Clamoring for ever more money, more funding, more “internationalism” with no real goals.
    Just saying,

    GoatGuy

  46. Perhaps a less tedious way to write that link: youtube /watch?v=dfY5ZQDzC5s (just strip the prefix, and replace the dot-com with a space).

    (Adding some extra text so that vuukle won’t count this as spam. This really is a circus of a commenting system, and we’re the ones jumping through hoops. It still wants more text. Ignore the brown fox jumping through hoops over the last dog.)

  47. If wishes were dimes, I’d be a rich man in my prime. Nice summary of the Much Ballyhooed future. Let’s see if it comes to pass. I think NOISE is going to scuttle the Universal Quantum Computer just as nonlinear instabilities have scuttled all hot-fusion efforts to date. Same issues, actually. GoatGuy

  48. I agree: it was the natural reduction to simplify everything as familiar click-bait … which seems to be the state of affairs these days across all media.

    I summarized it best in reply to WarrenTheApe. Go see…
    GoatGuy

  49. Couldn’t have said it better.

    ITER has become what virtually any relatively conservatıve Big Science (somewhat skeptical) analyst feared. A giant magnet (ahem) for near-limitless international Big Government Money spending. Spending on most anything that substantiates continued spending, stretching already picayune “lofty” science goals, and keeping “the patient” on life support for as long as it garners funding.

    “So, Joe… we have this big computer”
    “Yah, Bob? We have this huge reactor mock up”
    “Well, we think it is fast-and-smart enough to help you solve your stability problems”
    “Yah, Bob? We sure could use one of those ourselves. Could we buy some time on yours?”
    “Sure, Joe. Just put together a preliminary proof-of-requirements report, form J–1362–2017°F …”
    “Great Bob! Be on your desk within a month. Wanna do a press release?”
    “Awesome, Joe. Put that friendlier-face on the much maligned DoD”
    “Right, Bob. I’ll get Terry and Sandy in marketing to give your team a call.”
    “Good seeing you, Joe.”

    Something like that.

    THIS IS NOT TO SAY that using fast realtime computing methods mightn’t tease out of the prodigious data some previously unknown body of triggering evidence giving rise to plasma instabilities. And perhaps those newly discovered plasma burbles might later be something which electric and magnetic field modulation could counteract before they “upset the applecart”. But somehow … this old goat thinks that requiring exaFLOPS of computing power, taking hundreds of megawatts of energy to keep powered up, is not a terribly “commercial” reality. And no, I’d not be inclined to fall down the Quantum Computing is Around The Corner rabbit hole either.

    Just saying.
    You hit the nail on the head.
    ITER is a “Big Government Festooning Junket” if there ever was one.
    Din’t start out that way.

    GoatGuy

  50. Since VUUKLE chops out links, you’ll have to edit out the spaces to play it:

    https :// www . Youtube . com / watch ? v = dfY5ZQDzC5s
    … 0:22 → 0:50
    … 1:13 → 1:45
    … 2:05 → 2:30
    … 2:37 → 2:43
    … 3:11 → 3:27
    … 5:37 → 5:54

    SpaceX
    $25,000,000 to $75,000,000 for one intercontinental ballistic SpaceX people-mover.
    $125 M – $75 M / ( 10 to 100 people ) → $1.2 to $7.5 million per person
    Geometric mean… √( 1.2 × 7.5 ) → 3 million.
    Regular “average” would be higher ($4.35 mil)

    You’re acting the role of the Rubber Duck, sir quisling.
    Quack, quack … quack … quack.

    GoatGuy

  51. At first I thought the “AI” part was just sloppy reporting. After reading the original link, it turns out they want to use deep learning to predict plasma disruptions. That may indeed be doable, and in no way requires passing the Turing test or anything like that.

  52. We re not that far away from a universal quantum computer that can pretty much simulate anything, in a reasonable time frame. When that time comes, and it will, the quantum computer will probably show us several ways of creating fusion with the technology that is currently available. Perhaps there is even a way of creating cold fusion, that is unlike anything we have ever tried.

  53. Some one may say that the Turng test has little to do with cracking nuclear fusione problems, which may be heavily reliant on raw calculations. Same as saying that not being able to crack the Tuin g means computers are not good in calculations
    Goat Guy should not be taken as the Bible, especially after he said that planes fly as they “push air down” or that it would take some $3M to use the Falcon BFR to fly one person from one side of the globe to the other

  54. AI is all about reducing a system to the bare minimum number of features necessary to create a reasonably accurate model of the the system. So can nuclear fusion in a tokomak be accurately modeled with anything other than fusion in a tokomak ? I don’t know. But just saying you are going to do it with AI and some hand waving is silly. We don’t yet have a general purpose AI because no one yet has developed a universal learning node, where you just add more nodes to scale with what you are modeling. We do have artificial neural networks that can be made to work after a lot of head scratching and hair pulling out. But as of now you can’t explain why one approach worked and another did not.

  55. Almost from the start it was clear that fission was the better approach. Fusion is difficult to achieve under the best of circumstances, building equipment that will survive decades while producing it is going to be crazy difficult.

    While fission is so easy it almost happens on its own.

    The obstacles to fusion are unsolved physics, the obstacles to fission are just material science. And fusion faces those problems, too, once the physics is solved.

    Finally, we’ve known for half a century how to successfully cause fusion with excellent breakeven to occur, in bombs. And for almost all of that time have known that power plants could be based off those bombs. And yet we have expended huge resources on far more dubious approaches to fusion. Why?

    It’s the best which was designated to be the enemy of the good enough fission, that’s my conclusion.

    Just imagine how well developed fission power would be today, if all the work that’s been expended on fusion had gone instead to improving fission. What a waste!

  56. “The expected cost of ITER has risen from US$5 billion to US$20 billion, and the timeline for operation at full power was moved from the original estimate of 2016 to 2027.”

    Duh! ITER is to science what the High Speed Train to Nowhwere is to California what ULA/SLS is to NASA what SOEs are to the Chinese Economy: Nothing but pork-barrel, feather-bedding, vote buying Big Science Grant Whoring 101, kiddies.

    But then again, there are Kool-Aid drinkers on here who will scream otherwise. 5..4..3..2..

    “ITER will not generate commercial power.
    It will have have the follow DEMO pilot reactors project to prove out fusion at commercial scale then those will be followed by commercial nuclear fusion tokomaks.

    Those reactors are currently projected to be football or baseball stadium sized machines.””

    Duh! Re-read what I said in the above…

Comments are closed.