If you ask which of the scenarios I think is most dangerous, though, I will give a different answer. In that form of the question, I regard a nuclear attack, terrorist-generated or otherwise, as the most threatening combination of likelihood and long-term damage to modern life today.

I think that in the immediate aftermath of a superhuman machine intelligence revealing itself, most people would feel very threatened but take solace in the thought that we can always pull the plug.

I believe is that we can forecast the "changing landscape of context," and thus get insight into when we are entering the danger zone of an X-event.

When the next big problem comes online, be it the Euro crisis, nuclear proliferation, an overstretched Internet, a killer flu, or any of the other possibilities I consider in X-Events, we will suffer a complexity overload.

Trying to solve the problem by creating more debt is analogous to trying to stop being an alcoholic by going on a bender down at the corner bar.

To function effectively, the system scientist must know a considerable amount about the natural world AND about mathematics, without being an expert in either field. This is clearly a prescription for career disaster in today's world of ultra-high specialization.

I do not believe that there is any person, method, or tool that can consistently and reliably predict specific human events, X- or otherwise.