To summarize very briefly: The risks of global catastrophe are greater and more numerous than is commonly supposed, and they are growing, probably rapidly. They are growing for several reasons: the increasing rate of technological advance—for a number of the catastrophic risks are created or exacerbated by science and its technological and industrial applications (including such humble ones as the internal combustion engine); the growth of the world economy and world population (both, in part, moreover, indirect consequences of technological progress); and the rise of apocalyptic global terrorism. And the risks are, to a degree, convergent or mutually reinforcing. For example, global warming contributes to loss of biodiversity, an asteroid collision could precipitate catastrophic global warming and cause mass extinctions, and cyberterrorism could be employed to facilitate terrorist attacks with weapons of mass destruction. Each catastrophic risk, being slight in a probabilistic sense (or seeming slight, because often the probability cannot be estimated even roughly) when the probability is computed over a relatively short time span, such as a year or even a decade, is difficult for people to take seriously. Apart from the psychological difficulty that people have in thinking in terms of probabilities rather than frequencies, frequencies normally provide a better grounding for estimating probabilities than theory does; frequent events generate information that enables probabilities to be confirmed or updated. The fact that there have been both nuclear attacks and, albeit on a very limited scale, bioterrorist attacks—which, however, resemble natural disease episodes, of which the human race has a long experience—has enabled the public to take these particular risks seriously. The general tendency, however, is to ignore the catastrophic risks, both individually and in the aggregate. Economic, political, and cultural factors, including the religious beliefs prevalent in the United States, reinforce the effect of cognitive factors (including information costs) in inducing neglect of such risks. The neglect is misguided. The expected costs of even very-low-probability events can be huge if the adverse consequences should the probability materialize are huge, or if the interval over which the probability is estimated is enlarged; the risk of a catastrophic collision with an asteroid is slight in the time span of a year, but not so slight in the time span of a hundred years.