> It's baffling to me that so many people seem unwilling to accept the idea that maybe, perhaps, we've learned a few lessons about building safe nuclear reactors over the last 50 years.
I would suggest that the real problem with nuclear energy is that we don't know how dangerous it is. And we have no means of estimating the danger that isn't obviously inadequate. You see, nuclear disasters don't happen because engineers are incapable of designing failsafes and containment buildings. Nuclear disasters happen because it's impossible for engineers to envision all the possible failure modes. The confluence of events that causes a meltdown is inevitably something the engineers didn't design for.
In an enlightening article [0] written after the Fukushima disaster, a physicist and expert on nuclear safety argues that
- severe accidents at nuclear reactors have occurred much more frequently than what risk-assessment models predicted;
- the probabilistic risk assessment method does a poor job of anticipating accidents in which a single event, such as a tsunami, causes failures in multiple safety systems; and
- catastrophic nuclear accidents are inevitable, because designers and risk modelers cannot envision all possible ways in which complex systems can fail.
In other words, everything we "know" about nuclear safety is wrong.
Your argument is too powerful. It proves engineering as a whole is impossible. It equally proves we should never build any sort of building, because it could fall down and kill people and we can never be 100% sure we've prevented it.
It's an irrational appeal to emotion dressed up in rational trappings.
> Your argument is too powerful. It proves engineering as a whole is impossible.
No. It's an argument that we have no good way of estimating the damage caused by rare, spectacular failures in complex systems. But thanks for the down-vote.
> It's an irrational appeal to emotion dressed up in rational trappings.
No. It's pointing out an epistemic hole, one that is essentially the reverse of the sunrise problem [0], that has been discussed by countless philosophers of science, probability theorists, and scientists across dozens of fields. I suggest you read the article I linked and some of the papers it cites rather than make a spectacle of your downright hurr-durr ignorance of the subject.
This comment has now been both down-voted and up-voted multiple times. Would one of you down-voters care to explain why you feel this point is unworthy of discussion?
A blog post [0] that was up-voted near the top of Hacker News last week made the same point in terms that apparently are more palatable to some readers. If you haven't read it, I encourage you to do so. Two of the academic papers/book chapters [1, 2] that the blogger cites specifically discuss risk assessment of nuclear power stations in the context I described.
All of this stuff is well known. It's why web-sites' uptimes are never anywhere near the five nines advertised and why we have "flash crashes" in our financial markets.
The focus of my original comment is not only on identifying and assigning probabilities to causes of failures that are inherently difficult to discern, but also on the inadequacy of metrics like expected value in quantifying the effect of rare but spectacular outcomes. Is it really not even worth discussing these problems with our risk-assessment methodologies when the price of failure is as high as it is in case of an uncontrolled meltdown of a nuclear reactor?
Two problems — even if it's much unsafer than we think, it's still very safe, and comparisons to other power technologies and the effects of climate change must be made.
Suppose nuclear power is ten or a hundred times more dangerous than we think it is. Then from the figures that have already been quoted show it is ten to a hundred times safer than coal power (in terms of excess deaths from production of energy.)
On the price of failure, the price of failure to deal with climate change in the low-likelihood scenarios is billions dead. In my opinion, any discussion of nuclear power and low likelihood risks requires being put against those numbers, since nuclear is one of the only options to avoid climate change. In contrast to the billions that will die in large-scale climate change, a modern reactor can have an uncontrolled meltdown 2 miles from my house for all I care. It'll be an expensive and annoying mess, but it's not going to kill billions.
My argument is not that nuclear energy is much less safe than other energy technologies. I am merely observing that (a) our understanding of the safety of nuclear power generation has been shown to be severely lacking and that (b) nuclear power generation has a very different distribution of adverse outcomes than do technologies without the potential for catastrophic failure. Together, those facts suggest that comparisons of the safety of nuclear power to that of other technologies are not very meaningful.
But since it seems most readers are actively disinterested in the details, I'll give up the argument.
> It's baffling to me that so many people seem unwilling to accept the idea that maybe, perhaps, we've learned a few lessons about building safe nuclear reactors over the last 50 years.
I would suggest that the real problem with nuclear energy is that we don't know how dangerous it is. And we have no means of estimating the danger that isn't obviously inadequate. You see, nuclear disasters don't happen because engineers are incapable of designing failsafes and containment buildings. Nuclear disasters happen because it's impossible for engineers to envision all the possible failure modes. The confluence of events that causes a meltdown is inevitably something the engineers didn't design for.
In an enlightening article [0] written after the Fukushima disaster, a physicist and expert on nuclear safety argues that
- severe accidents at nuclear reactors have occurred much more frequently than what risk-assessment models predicted;
- the probabilistic risk assessment method does a poor job of anticipating accidents in which a single event, such as a tsunami, causes failures in multiple safety systems; and
- catastrophic nuclear accidents are inevitable, because designers and risk modelers cannot envision all possible ways in which complex systems can fail.
In other words, everything we "know" about nuclear safety is wrong.
0. http://thebulletin.org/web-edition/features/beyond-our-imagi...