For further reading:

Two Brazilian physicists weigh in on Hawking radiation (or lack thereof):

http://arxiv.org/PS_cache/arxiv/pdf/...808.2631v1.pdf

And here is the paper from the ArXiv on catastrophic risk analysis that makes the point I've been trying to make all along much better than me:

A critical look at risk assessments for global catastrophes

Someone was asking where I was when they first lit off RHIC. Actually I was drawing a paycheck at the time from the Department of Physics at CSU (they wanted a philosopher of science to help TA an undergraduate class). One of the philosophers, Dan Lyons was freaked out by Rees' estimate that there was a 1 in 50,000,000 chance that RHIC would destroy the world, and so Dan wrote letters to Brookhaven (I have a copy of the letter they sent him somewhere). And so naturally, RHIC came up for discussion with my boss, physicist Marty Gelfand. Marty didn't seem to think a 1 in 50,000,000 chance was a big deal because it was such a small probability that it was not worth worrying about. Nothing I could say would convince him otherwise.

But the paper above by Kent makes the case clear. I've been criticized for trying to tote up the value of the world in USD. Well, Kent merely uses human lives. Thus each life has a unit value of 1. Thus if the population is 6 billion, then the value of the Earth is 6 billion (neglecting of course nonhuman values as well as the value of future lives). Thus "the expectation value of the number of deaths" for RHIC (using Rees's estimate of 5 x 10^-7) was 3,000 deaths.

(Now one could say that since we lit off RHIC and nothing happened, then that shows that the risk was justified. But I think that would be the wrong way to look at it. Imagine you were a passenger in a car I was driving and I said "Watch this!" while I drove around a blind corner in the wrong lane. We survive--and you're mad as hell. "But," I say, "nothing happened!")

Well, what would be a reasonable maximum risk of global catastrophe that we should accept? Kent answers that question by looking at risk regulation in other areas like radiation risk management. If radiation minimization standards were to apply to potential existential threats, then the probability of catastrophe (p_catastrophe) must be less than 10^-15--and that's not taking into account the value of nonhuman nature and future generations. And according to John Rawl's veil of ignorance, our generation ought to be no more privileged than any future generation, and that therefore, we ought not to discount the future.

How many more orders of magnitude should we add? That is a question I wish you all would address. Given that Homo sapiens as a species is about 100,000 years old, (the average lifetime of a mammalian species is about 2 million years), I would say we have a responsibility to think at least 100,000 years into the future (i.e., our ancestors for the past 100,000 years never did anything rash enough to cause our extinction, therefore we should not do anything rash enough to cause future extinctions at least 100,000 years into the future.) So if we take an average life expectancy of 100 years, then that's at least 3 more orders of magnitude that one should add; thus, the acceptable p_catastrophe should be less than 10^-18. If we really want to look 1 billion years into the future, as cjameshuff occasionally likes to, then we would have to add seven orders of magnitude such that p_catastrophe would have to be less than 10^-22--lower than Neverfly's estimate of p_catastrophe at 10^-21. But I could accept p_catastrophe's of less than 10^-18 as counting as "reasonable certainty" that nothing catastrophic will happen.

However, I must say that my own estimates of p_catastrophe are much greater than 10^-18. What is the CERN estimate of p_catastrophe? The fact is we don't know. It wouldn't surprise me if some CERN scientists have in fact calculated p_catastrophe. Physicists make such bets all the time. Steven Hawking recently bet $100 USD that the Higg's boson will not be found at the LHC. That is, he figures p_~HB = 50%. So I wouldn't be surprised if some CERN scientists have calculated a p_catastrophe for the LHC. But if so, such estimates have not been made public. And why should CERN choose to make public a quantitative estimate of p_catastrophe? After all, they're not required to. As a result, we the public have to make do with p_catastrophe = "really, really low".

If there is a lesson in this whole LHC saga, I think that it is that government environmental regulations for dealing with small probability, but potentially catastrophic, global risks are inadequate because such regulations do not exist. As Kent says, we need to reach a general agreement on what probabilities of global catastrophe are acceptable. These should be determined ahead of time, and projects such as the LHC should be required to provide a quantitative estimates of p_catastrophe, and these estimates should be verified by government scientists who are not personally invested in the project. If the developers of the project cannot demonstrate the p_catastrophe for the proposed project is not less than the maximum acceptable p_catastrophe, then the project should not be permitted.