I think this claim is exceptionally hard to support. It’s impossible to copy anything from a human today, except things we can write down, and I don’t see that limitation being lifted in the near future. Human brains are very much unlike semiconductor-based computers, and inasmuch as computers have been built to emulate human behavior, that emulation has come at tremendous computational cost.

Watson, for example, was a room full of computers. 8 full racks with support equipment. And that was a computer designed to perform one very specific language task (and it was allowed to cheat in that it did not have to visually read or audibly understand anything like its human counterparts; the text was fed to it in pre-digested computer compatible form). Even if we imagine Watson being reduced to the size of a Speak & Spell (“The TI Answerman 1000!”), it’s still a very specific task.

You could certainly assemble enough of these very specific task systems to get fairly human-ish capabilities, but it would be nothing like a copy of a human brain.

Human brains are full of little analog states, highly interconnected with highly “parallel” implementation. Many tasks seem to light up the whole brain — tasks that, on a computer, would be performed by a specific subcircuit of the CPU with binary precision. Getting anything parallel to work right in a computer — at least where the problem doesn’t easily break down into pieces — is still a major computer science research area.

More importantly, there is no way to read the state of a human brain, and it’s not clear that there ever will be sufficient detection precision to do so. So even if you could create a perfect brain-analog computer, initializing it with somebody’s current brain contents is pure science fiction. How would you read the states of billions of neurons?

If you could make a sufficiently power brain-analog computer, you might be able to teach it from a first principles framework. But that’s very different than making a copy.

First, the rich are not in a position to consume more and invest less – perhaps Mr. Economist Bryan Caplan was sick that day when they covered marginal propensity to consume. Second, we’ve played with tax rates in the USA and have not seen what he’s talking about.

Third – capital flight? Why doesn’t he just say ‘going Galt’?

Robotopia!

Even if AI equivalent to human intelligence were necessary and desireable, how does your system work, really? Most machines will be run on electricity, so either they can be equiped with systems that draw energy from the environment (solar mainly) or will need to be constantly or periodically connected to a power source. We already have a centralized system of power generation, and there is no reason for that to change. Why then would robots gain control over that most fundamental of their needs? The economic system of 1600 you described existed only because land was fundamental to the maintenance of agriculture, and this was key because people have to eat and drink above all else (breathing is assumed given air’s nature) This system does not work for this robotic future. Their basic needs will come from large, established centralized systems or completely decentralized systems in which land ownership is immaterial.

And what about human needs? Having robots does not remove the basic human need for food – this means land. Will every human own their own plot of land that their robots will farm? I don’t see how we could go back to such a system, so the basic needs for most humans (agriculturally productive land) will still be controlled by only a minority of folks, as it is today. Currently, what most humans do to have access to this product of a few is conduct work for wages that they can then trade for food. What happens when most humans don’t have wages to earn? How do they get food?