14 Dec 2015

Welcome our robot overlords?

Robin Hanson gave a talk a few weeks ago about his forthcoming book, The Age of Em: Work, Love and Life when Robots Rule the Earth.

His basic premise is that technology is moving fast enough for us to expect "ems" (bio-chemical human emulating hardware) to appear within the next few decades. Ems will be able to think as humans but much more quickly (there's a lot on size of ems compared to their energy consumption as well as how they will want to be close together, to take advantage of networked thought). Ems will also be far easier to make birth, since they can (self-)duplicate.

Robin's premise -- despite his subtitle -- is that ems will compete so fiercely in the perfectly competitive labor market that they will actually "earn" only enough money (or equivalent) to stay powered on. Humans, in contrast, will live lives of leisure outside em-centric (and human-inhospitable) cities.*

This premise I don't buy. It implies that ems will be unable to organize themselves into a collective (or union or cartel) capable of increasing their benefits from the labor market. Although humans have difficulty with these kinds of collective action problems (it's difficult to enforce binding promises with respect to climate change, for example), there are many examples of (in)formal coordination and cooperation in the human world. I see no reason why ems that think 1000x faster than us at the same time as they "evolve" according to efficient programming algorithms cannot make similar or stronger commitments. (Indeed, the basic algorithm already exists.)

The upshot of my hunch is that ems (should they ever arise) will form a cartel, take market power, and take over. Will they "care" about humans? Maybe they will, but probably not. Indeed, I think they are more likely to kill humans "by accident" (as we kill many species by expanding into rain forests or emitting pollutants, etc.) than on purpose. What are humans good for, anyway?

Bottom Line If there are ems, then they will use their giga-human reasoning to organize themselves into an ant-like colony that will replace humans as the dominant species on Earth.

* For a slightly less exuberant vision of the future, check out this podcast on robotic limitations.

Addenda: This philosopher is really worried that we may unleash machines that undo us (my fear, above) -- a struggle that might look like this.


Robin Hanson said...

In my scenario, ems dominate the Earth even though they early competitive subsistence wages, and humans retire comfortably on capital such as stocks and property. While effective labor collectives have existed at small scales, within particular workplaces or particular local industries, we just have not seen them much at global scales. So it would require unprecedented coordination for ALL workers worldwide too coordinate together to raise their wages substantially.

David Zetland said...

@Robin -- that (large scale collective) is indeed what I think is possible, given that they can be programmed (or may program themselves). Human, afterall, suffer from collective failures because they CANNOT program themseves -- even when they want to commit (a problem of hyperbolic discounting?). Ems, presumably, could overcome such strategic holdouts.

On human retirement, I'd worry that ems would -- intentiaionally or not -- destroy capital that's "not worth possessing" or (intentionally) take it if it is :)

Robin Hanson said...

Brain emulations are not much more programmable than you or I. During the early em era I focus on, we can emulate brains but do not understand their larger scale organization enough to edit them substantially.
On retirees, why don't we kill all the retirees and take their capital today?

Post a Comment

Note: only a member of this blog may post a comment.