I'm struck by the vastly incompatible views of the future he seems to be holding simultaneously. At the beginning he says, "economy will double every two weeks," and "if you save a little the investment money will make you rich from the economy doubling" coupled with "the entire EM economy could be located all in one small Scandinavian country near where uploading is first created," and "the population 10 trillion cities will consume huge amounts of resources, destroy the environment, and eventually consume the globe." How does that count as anything except a hard takeoff that leaves humanity in the dust, and how do you plan to cash out your investment from the immortal Em society when they decide not to honor your 10^26% compound interest?

I was left with the impression that a lot of the ideas were made because he wanted them to be real first, then were rationalized how it would work later. Some of the things discussed seem especially incoherent. His whole section on Em socialization seems odd when you can just duplicate friends at any time. In fact, the entire talk seems predicated entirely on having near-perfect mind theft security which is covered at the very end in a single slide.

There's also a bit about corporations at ~37 mins. A company with 256 line workers that organizes in 4 man teams needs with 64 line bosses, 16 higher bosses, 4 VPs, and one executive. That's 5 levels of communication which can cause all sorts of problems, not to mention the cost in ~25% of the company being managers. The solution he describes is a single, time accelerated boss running at ~21x speed able to coordinate all 64 teams and keeping the entire company running as one entity. (Note that uber-boss is still taking up 25% of the company's em time.)

The issue I have with this is that it ignores the entire advantages of emulation discussed earlier in the talk, namely duplication of workers. If you can duplicate workers effectively infinitely (eg plumber example) and you can create and destroy workers every 10 minutes (eg computer programmer example), then you don't need an organization at all. Parallelizable tasks can all be solved whenever (again, plumber), and all the work that remains will be in serial. Ems will be tasked with jobs on their instantiation, and will return their results to whoever needed it when their task is completed (which could just be another worker). This brings corporations outside of the realm of bosses and economists and pulls them into the realm of computer scientists. Corporations will look a lot less like pyramids and a lot more like computer programs.

I feel like I'm gonna lose some karma on this comment, but... yeah. Hanson normally strikes me as very insightful but this talk just felt like a mess.