In this week's Rule Breaker Investing podcast, Motley Fool co-founder David Gardner has invited a special guest: Robin Hanson, a George Mason University economics professor and research associate at Oxford's Future of Humanity Institute. Hanson has master's degrees in physics and philosophy and a Ph.D. in social science, and he previously worked on A.I. at Lockheed Martin and NASA. His recently published book is The Age of Em: Work, Love and Life when Robots Rule the Earth.
Hanson joins us this week to discuss a broad topic: the future. His key prediction is that digital brain emulation will essentially create a computer model that can replicate human-like processes and actions. So if an emulation of you could do your job cheaper than you, then where does that leave us?
This video was recorded on Sept. 14, 2016.
There's a lot we don't know about the future, so my book is premised on assuming that a particular kind of robot is the first kind of robot to be as smart as humans broadly, all across the board. I don't know if that's true, but it seems worth considering a whole bunch of different scenarios and working them out. The future is important enough to have a hundred different books working out a hundred different scenarios that have only a 1% chance. I'm happy to think my book at least meets that standard.
So a particular kind of robot, and it's called a "brain emulation." That's why it's called The Age of Em. "Em" is short for emulation.
Yes. And the idea of a brain emulation is you take particular human brains, you scan them, and find spatial and chemical detail. You have models on how each of the kinds of cells works and you make a big computer model that works like the original brain in terms of input/output signals. And that means if you would hook it up with hands, eye, ears, [and] mouth, you could talk to it. It would talk back. You might ask it to do a job, and it might do it. And if it were cheaper than humans, everything would change.
Now, this route of artificial intelligence plausibly might happen roughly sometime in the next century, which is why I think it's an interesting scenario. Actually, at the rate we've been going in the ordinary kind of artificial intelligence, it would take two to four centuries before we reach human-level abilities there, all across the board.
But what has happened typically (and certainly Ray Kurzweil and other futurists, but particularly Ray comes to mind), just running Moore's law, [is] making assumptions that double, double, double. Maybe it takes a year or two to do those doubles, and it doesn't look like much when you go from one to eight.
Yes, that's a standard story, but honestly my guess is that there's a log-normal distribution of the cutoff for jobs in terms of automation. So in the last 50 years, we've seen exponential growth continuing where we double capacities every two years, but we have not seen exponential displacement of human jobs. We've seen relatively steady displacement of human jobs. That suggests there's this wide distribution of how much computing power is required to displace any particular job, and there's a long way ahead. We need to have a thousand, a million times more computing power to displace more jobs farther up the ladder.