Yes and no. For safety of narrow AI systems, yeah, there's a lot of scope for research, and that's what your first link gets at. But for AGI (which is what Te

Yes and no. For safety of narrow AI systems, yeah, there's a lot of scope for research, and that's what your first link gets at. But for AGI (which is what Tegmark talks about), there's no good way to get a handle on safety yet (other than working towards figuring out AGI). As for MIRI's agenda, I don't buy that it will help with AGI safety at all. There are a variety of reasons for that, some of which are discussed in the piece I linked above.