I say this because of the unlikelihood of actually obtaining resources for such an intelligence, which seems unlikely. Although it could attempt to hack for bitcoins or work for money, it would have a hard time actually buying something and gaining anyone's trust. However, if such a superintelligence does get access to resources, it is actually dangerous because the alignment problem is real, so I would put this at around 10 to 20%. You can give any arguments in the comments, and I'll read them and change my percentage accordingly. The market resolves to my final estimate of P(doom) of AI superintelligence.
1,000My p(doom) is lower, but for different reasons. Why would it have to gain trust? As I see it, ASI would immediately be utilized in the economy. This isn't mythos-level "marginally improve cybersecurity defenses before our adversaries attack", it is radically upend human society, advance medical science, improve upon all knowledge-work.
@NivlacM interesting. The superintelligence refers to the intelligence itself. So, gathering resources is a task that it has to perform, which requires physical impact, unlike the intelligence.