‘GenAI for Nuclear Licensing’
https://www.youtube.com/watch?v=PW9lusiwMz8&list=UU9rJrMVgcXTfa8xuMnbhAEA - video
https://pivottoai.libsyn.com/20251118-vibe-nuclear-lets-use-ai-shortcuts-on-reactor-safety - podcast
time: 6 min 24 sec
It’s not even trying to solve the right problem. In the US, the NRC has given out licenses for new reactors. They’re sitting there without the funding needed to go forward.
I have no doubt that licensing is a long process. It should be. That’s how we keep fission power safe. But the more fundamental reason they’re not getting built is because they reliably blow their budget and schedule.
Hell yeah.
Nuclear energy isnt a technical problem, it’s a human problem. Specifically, the real expense in US nuclear construction is that there are only a handful of contractors who have the tribal knowledge to actually do nuclear construction e.g. pour concrete, install old-fashioned non-networked electrcal control systems, big switchgear, pipefitting, startup V&V, an so on.
They’ll all gladly monkeywrench, slow walk, and re-work every step because they know there’s no real competition for fleet-wide contracts, and no one from the CEOs to the craft on the ground want the job to end, so you get it decades late or not at all.
One more piece of evidence that prompt fondlers are not serious people.
Source: am person of nuclear
Yeah, even before the techbros showed up, there was this industry push to try to convince people that regulation was the problem. If we loosened the bolts just 10%, everything would work out, they think. Attacking the “linear no threshold model” seems to be the latest strategy.
It’s almost like there’s a reflexive need to blame government regulation on all the problems.
Even ignoring AI datacenter builds, we still need clean energy. I would be all for nuclear fission if it were at all economically viable. It just isn’t.
yeah, even the green case for nuclear - which has been around for a long time - falters on wind and solar with battery just being hilariously cheaper. At this point the funding problem is interconnects.
Of all the things that will never happen, this will never happen the most.
Yeah. Currently the center-right government here in Sweden are pushing nuclear, because the one thing they hate more than brown people is green people. But the industry actually tasked with building and running the plants have extracted both credit guarantees (== taxpayer money) and a legally mandated minimum price on electricity - remember, the entire populist narrative of nuclear is that it will bring cheap power! Plus they’re unlikely to start building unless there’s a stable parliamentary majority to cement the deal.
OK, so far, so good (nuclear policy in Sweden is colored by the fractious debate on ending nuclear power in the late 70s, but the principals of that debate (boomers) are finally dying off). But when the shovel hits the ground, you need a site to build it on. And the neighbors of said site are generally of the opinion that a fucking power pylon is an infringement on their god-given right to property resale price increase. Imagine a concrete box of eldrich power blocking the skyline, and imagine a population of people who know how to pull the levers of obstruction and have had decades of practice doing it. Fun times ahead.
If the AI bros get these things built before the bubble pops, we should expect bad designs, not allowing for local conditions, setting up the reactor operators for ridiculous errors, and lots of nuclear accidents. Hopefully not very big ones. Cross fingers!
If and when those accidents start happening, its going to set back adoption of nuclear power by years, if not decades - especially if there’s an incompetent response to those accidents (which, considering Starmer and Trump are in charge, is worryingly likely)
One of the other important things about the nuclear regulation process is that it makes sure the local people are involved. You can’t skip that step either. If you just run roughshod over the locals for the sake of AI, you’ve already got people in the streets protesting AI.
The government will almost certainly try to just bypass the activists. But remember: anti-nuclear activists have decades of experience at this. So I’m sure it’ll go great all round.
Anti-nuclear activists are going to have a field day with this, aren’t they?
have you seen how much time it takes to built single NPP? openai will be a smoking crater well before site for the first plant will get selected. then you have a backlog for turbines and reactors
Given the state of renewables and energy storage, this feels a lot like the final opportunity for nuclear power in its current state to actually do anything at all, and the “move fast and break things” crowd have no idea about building physical things more complex than a datacentre which honestly, isn’t that challenging in comparison.
openai will be a smoking crater well before site for the first plant will get selected
Other things that might not last that long include the government of the country in which you’re trying to build massive piece of infrastructure that represents a significant ongoing maintenance burden and risk.
This could set back nuclear a lot actually. Get ready for anti-nuclear activists using this in the near future as an excuse to push coal and oil as “safer” alternatives.
@ZILtoid1991 Spoiler: it won’t happen, it’s basically more marketing bullshit from the folks trying to make bank off the AI bubble.
Next they’re going to tout AI-powered diaper-changing robots for nursing homes and maternity wards. Equal likelihood of success, of course.
have you ever met actual anti nuclear activists
