I’m not sure specifically about IndoPak nuclear war, but I think the general drift of this post is important—lots of Xrisk comes from factors we haven’t thought extensively about (I don’t think this particular area has come up on LW before).
There are two models for thinking about Xrisk. In the first model, which I associate with SIAI, most Xrisk comes from a small number of enumerable threats: uFAI, asteroid impact, etc. In this model, to prevent extinction, one should consider each threat in turn and design measures to protect against each.
In the second model, which I personally favor, the majority of Xrisk is distributed among a large number of threats. Each individual threat is highly improbable*, but because such threats are so numerous, net Xrisk is significant. In this model, it is impractical to design specialized defenses against specific threats. Instead, one should design countermeasures that protect against entire categories of threats.
* More precisely, the events may be probable but the likelihood of the event causing existential disaster is very low. IndoPak nuclear was is moderately probable, but it seems highly unlikely that such a war could pose a serious threat to human survival.
I’m not sure specifically about IndoPak nuclear war, but I think the general drift of this post is important—lots of Xrisk comes from factors we haven’t thought extensively about (I don’t think this particular area has come up on LW before).
There are two models for thinking about Xrisk. In the first model, which I associate with SIAI, most Xrisk comes from a small number of enumerable threats: uFAI, asteroid impact, etc. In this model, to prevent extinction, one should consider each threat in turn and design measures to protect against each.
In the second model, which I personally favor, the majority of Xrisk is distributed among a large number of threats. Each individual threat is highly improbable*, but because such threats are so numerous, net Xrisk is significant. In this model, it is impractical to design specialized defenses against specific threats. Instead, one should design countermeasures that protect against entire categories of threats.
* More precisely, the events may be probable but the likelihood of the event causing existential disaster is very low. IndoPak nuclear was is moderately probable, but it seems highly unlikely that such a war could pose a serious threat to human survival.