AI and the Nuclear Caution
Overly risk-averse policies hurt us in the long run, even for risky technologies
Is AI The Bomb or the Printing Press?
In my previous article “AI Risks and AI Regulations”, I made the argument that premature and excessive AI regulation will “slow down the development of highly beneficial AI technology,” and that instead we should instead “support the freedom to build and use AI technologies” so we can maximize our benefit from AI.
The counterpoint argument is that a technology as powerful as AI is inherently risky and therefore we should regulate it for safety, not freedom of use. When such arguments arise, comparisons with a technology that is inherently risky, in particular nuclear technology is made. A rhetorical question by Senator Hawley last month also posed the comparison:
My question is what kind of innovation is it going to be? Will it be like the printing press … or like the atom bomb? - Senator Josh Hawley
I am going to dive and discuss nuclear technology in this article to make the point: Not only should we not regulate AI like nuclear technology, but we shouldn’t even have treated nuclear energy like we have. We have led from fear and over-caution on nuclear energy, and we’ve paid a stiff price. AI is an informational technology more like the printing press than the atom bomb, but even if AI were like nuclear technology, regulating AI from a position of fear is the wrong prescription.
As with AI, nuclear energy policy can be viewed as a debate between Optimists and Doomers. Optimists once predicted nuclear energy that ‘was too cheap to meter,’ and they now hail it as a reliable, zero-emission clean energy source. Doomers have lamented nuclear energy’s dangers, including the risk of accidents, the disposal of radioactive waste, and the risk of nuclear weapons proliferation.
These fears have led some to call for stricter regulation or even phasing out of nuclear energy, which has influenced nuclear energy policy and hurt the industry. Three specific nuclear energy policy decisions stand out:
The ban on nuclear fuel reprocessing cycle
Government regulations on nuclear power plant construction
Defunding nuclear research reactors
Ban on Nuclear Fuel Recycling
The ban on the closed nuclear fuel cycle was a decision by President Jimmy Carter, based on fear that reprocessing nuclear fuel as a technology would add to nuclear proliferation risks. While this fear is not unfounded, it’s hypothetical; in reality, stolen used or reprocessed commercial nuclear fuel never led to nuclear proliferation.
The lack of reprocessing nuclear fuel closes off promising technologies to handle used nuclear fuel. Without recycling, used nuclear fuel becomes ‘nuclear waste’ that remains radioactive for tens of thousands of years. It is over the long-term safer and much more environmentally friendly to reprocess and recycle nuclear fuel, using ‘actinide burning’ nuclear power reactors to completely burn it up. This could cut the level and impact of nuclear waste one hundred-fold. Closing off this technology added to environmental and cost burdens of nuclear power unnecessarily.
NRC regulation on plant construction
While nuclear power has had a remarkably safe operating record, it has been punctuated by three serious accidents that have shaken the nuclear power industry: Three Mile Island, Chernobyl, and Fukushima.
The first and only US-based accident was the Three Mile Island (TMI) accident in the US in 1979. An operator error led to a reactor core meltdown that caused a panic and evacuation of thousands of people. In the end, nobody was hurt by it, but a reactor was destroyed. The TMI accident put the nuclear power industry under heightened scrutiny, led to increased public opposition to nuclear power, and made the siting of power plants in some localities difficult to impossible.
The Nuclear Regulatory Commission (NRC) reacted with elevated regulations on plant construction. These heightened safety considerations imposed bureaucratic red tape on construction that led to huge construction costs on nuclear power plants. Costs cascaded and cost overruns plagued nuclear power plant construction after TMI, with some plant construction ending in expensive failure.
As a result, the industry that had built over 100 nuclear power reactors in the 1960s and 1970s slowed new nuclear power plant building to a crawl as new nuclear power plants became cost-prohibitive to build. Nuclear power plant construction fizzled out in the US and has yet to come back.
The irony is that despite a few spectacular accidents, the nuclear power industry has had a remarkably safe operating record since then, with the net result that nuclear energy is the safest form of energy we have. It’s so safe, we cannot afford to build it.
Freezing nuclear power research
The third decision that froze nuclear power research progress was the decision to curtail nuclear power research. In 1994, the Clinton administration decided to shut down the Integral Fast Reactor (IFR) program at Argonne National Laboratory. The IFR was an experimental breeder reactor which could address the nuclear waste issue because it could produce as much fuel as it consumed.
Since then, there has been no US Government funded effort to build a working prototype reactor, resulting in stunted progress on nuclear energy technology - a ‘nuclear energy technology winter.’ Despite this, there are many advanced reactor concepts on the drawing board, and various energy startups such as TerraPower have proposed new next-generation concepts that offer a path forward. We just need to build the prototype reactors to prove they work.
Our Possibly Bright Nuclear Future
These three decisions - banning nuclear fuel recycling, over-regulating nuclear power plant construction, and de-funding nuclear energy research reactors - hampered and stalled nuclear energy for the past four decades.
If we reversed these three decisions, we could enable new research and development of nuclear technology, un-freeze nuclear power development, allow for a nuclear fuel cycle that is orders of magnitude more efficient and environmentally-friendly, and lower cost burdens on building and operating nuclear power plants.
This would kick off a new era of safer, cleaner and cheaper 21st century nuclear power technology. The new nuclear energy era would be based on more advanced nuclear reactor technology than the 1970s-era light water reactors, specifically molten-salt reactors.
Molten salt reactor cores have many advantages: They are liquid at operating temperatures and not pressurized, which makes stable, melt-down-proof and inherently safe in operation. Their high operating temperatures increase thermal efficiency, making more power for less cost. They are less expensive to operate. This form of nuclear energy would be the cheapest, safest and cleanest form of energy we could produce today.
The Lessons for AI
We have nothing to fear but Fear itself. - President Franklin Delano Roosevelt
Our experience with nuclear power is a lesson in how fear led to stunting energy development and technology progress, hurting our access to cheaper, better and safer energy.
The lesson for AI is a cautionary tale: Risk-averse policies that close off technology and development hurt us in the long run, even in industries and technologies that are inherently risky. Even when a technology comes with risks, regulating based on excessive caution yields worse outcomes than policies that encourage free development of new technologies.
If we treat AI as a dangerous technology like nuclear weapons, then we will live in a world where access to AI is limited to certain countries and certain powers, where the ability to take advantage and benefit from AI is limited. The utility of AI will be stunted and access unequal. This is not the world we would want for ourselves and our children if the alternative is advancing AI technologies available to all.
But AI technology is not like nuclear weapons, it is more like the internet or the printing press. The broadest access to AI will yield the broadest benefit to mankind. AI as a technology can help us make wiser and better decisions. We can hope that wiser decision-making in technology policy itself will be one outcome.
Postscript
The podcast “No Priors” hosted by Elad Gil and Sarah Guo covers artificial intelligence, and I highly recommend it. Elad Gil in a June 8 podcast made comparisons between AI regulation proposals and the problems with nuclear power industry regulation. I have for many years known about the problems with nuclear energy mis-regulation and the price we’ve paid for bad policy. Elad’s comments inspired me to flesh out a fuller explanation of this issue and its lessons for AI. Thanks and tip o’ the hat to Elad.