I think we don't know what we don't know. It's not hard to imagine scenarios why which AI could be weaponized in a winner-take-all scenario.

I've noticed some in this corner of the internet tend to take the point of view that decentralized protocols are the ultimate defense against these downsides, and I find that disturbingly naive. I do not share the view that the nation state is now effectively obsolete and we should accede to purely bottoms up, decentralized and communitarian approaches to such an asymmetric threat.

I tend to take the point of view that the dangers will very much manifest in the physical realm, and sometime in the next decade or two, we will likely touch the fire. The question is whether we will be able to put that fire out or not. I don't know how to quantify the risks there.