[mlyle]

I'm not in favor of a full-on precautionary principle for everything.

But if you take on too many 1 in 1000 or 1 in 10000 risks, eventually things go badly.

Russian roulette or reckless driving looks safe-- for awhile.

Note the person who really advanced the view of "igniting the atmosphere" with fission was Teller back in 1942, who was about the most pro-nuclear scientist you can find. They didn't just shrug and push the button: Bethe and others did a lot of math to conclude that it was exceptionally unlikely.

(Was it too much of a risk, given that we didn't have much experimental data about nitrogen-nitrogen cross-sections? Probably not, but we can't conclude that it was a reasonable risk purely on the basis of "we didn't die.")