AI risk arguments about real-world physics and self-replication

One post challenges AI-risk skeptics whose safety argument depends on future superintelligent models being unable to interact with and research new physics in the real world. Quoting that argument, another post raises a speculative concern: to the extent we understand AI improving itself and nanoscale materials or robots that replicate themselves, it argues, we have ingredients for self-motivated, global expansion of “Something Besides Us.” The author explicitly allows that extinction by “grey goo” may not happen.