michael_nielsen

1910: “Planes are getting much better. I worry they’ll accidentally go into orbit” Aircraft people: “well, that might be possible one day, but not with canvas & props” 2017: “AI is getting much better. I worry it might accidentally turn into general AI” AI people: “well…

I believe recent analyses show very little agreement amongst researchers on this topic. @Miles_Brundage is that right? What's the best study to look at re researchers expectations on AI threats?

there are others but generally smaller and more biased/problematic in various ways (not that this one's perfect, but it's more of a serious effort than most such surveys)

Interesting that it's 21% of NIPS/ICML 2015 authors. That's actually not a terrible sample of what I'd consider the professional ML/AI community. Hard to argue for anything other than, as you say, "there's little agreement in the research community".