Michael,

Solidly presented arguments about the dangers here. Cudos.

You made ‘party-line’ assumption statement, that unfortunately I think is not right: ``The possibility of AGI (artificial general intelligence) manifesting in the 21st century is low’’

80 years is a **LONG** time in technological terms. I know it seems we are quite far from AGI now, but the thing we must remember is that most of AGI is developed by the AGI itself. Even as we lay the last tech-brick needed, it will appear we are very far from AGI even as we already built the seed that will become it.

If we graphed the max distance humanity had risen above the earth’s surface over time, no extrapolation from that data could have predicted our rapid progress to the moon. It simply was not supported observed prior progress. But this is how thing happen with jumps — in order to analyze the time frame of the jump one needs to analyze the required ingredients. Alas I think we have built far more of them, than is generally recognized (even within my field).

p.s. I was a DARPA PM driving precursor tech for AGI, so I have thought quite a bit about this topic. I have thought of writing a book that lays out the argument about how AGI will come to be. Probably the most important thing I could be doing with my hours. but difficult to get paid for such work.