Bryan Caplan  

Robin Hanson's The Age of Em: A Succinct Assessment

Rushkoff's nostalgia... Cornucopia...
I've criticized Robin Hanson's Age of Em for being needlessly confusing.  To avoid being subject to the same charge, here is my succinct assessment.

1. Robin Hanson is a brilliant, delightful thinker, and I'm glad he wrote the book.

2. Futurism - especially futurism informed by social science - deserves much higher status and vastly more intellectual attention.

3. Artificial intelligence will be important in the future.

4. While it's possible to bypass philosophy of mind and simply describe the future role of AI, Robin's whole analysis tacitly assumes an extreme version of "ems are just as human as you or me."  If he were really agnostic, roughly 50% of the book would have focused on the lives of biological humans during The Age of Em.  The true share is more like 2%.

5. In practice, the difference between ems and generalized AIs will be modest, because we'll pre-select robot-like humans to emulate.  Talking about ems' love lives and religiosity is silly.

6. Robin describes the Age of Em in lurid language, then wonders why readers are afraid.  How does he expect readers to react when he tells them that "ordinary humans" will be "sidelined" and "earn zero wages," while the "vast majority of people will live at the subsistence level"?

7. In plain English, however, Robin's description of his scenario is very bright.  Biological humans will enjoy immense prosperity.  If simulated humans are conscious, their lives will be hard, if not hellish.  But why would mere simulations be conscious, anyway?

8. As long as AIs are psychologically robot-like, biological humans will remain in charge of politics and business.  Biological humans won't be dominated, expropriated, or exterminated by their own creations.

9. However, if AIs are psychologically human-like, ems will probably do terrible things to the first generation of biological humans to meet them.  Since the ems subjectively experience years in a single objective day, a small risk of em-human conflict per em generation yields a very high risk of em-human conflict per human generation.  Over time, moreover, this risk is likely to rise, because the ems would, within a few of their own generations, develop a radically separate identity and social network, creating preconditions for ugly - and plausibly genocidal - group conflict.  Remember: In Robin's scenario, the ems vastly outnumber the humans, even though the humans have the lions' share of the wealth.  And the humans no longer contribute anything to the global economy; they're true rentiers.

10. No matter what happens with AI, the global economy will never double in size annually, much less monthly.  There are too many political, economic, technological, and social bottlenecks.  In any case, the extraordinary claim that the economy will double on a monthly basis requires extraordinary evidence that Robin definitely does not possess.

COMMENTS (4 to date)
Silas Barta writes:

So you take the extreme Penrose/Searle position that computers can never be conscious? If so, what do you think makes them unable to be conscious like humans, in response to e.g. Scott Aaronson?

michael pettengill writes:
No matter what happens with AI, the global economy will never double in size annually, much less monthly
Define AIs and EMs buying from each other using bitcoin as production, eg, buying poetry, and gdp can triple easily every day!

You can't be claiming that gdp is in any way connected to labor cost?!?

In the perfect economy, all labor costs are eliminated! Right???

Elias Hakansson writes:

I don't see what is so inconceivable about an economy doubling monthly in an age of em. The capital necessary to ensure subsistence of an em is simply hardware, electricity and real estate connected to a communication grid. A monthly doubling time for these things seems perfectly reasonable. As long as the doubling time of em software doesn't exceed a month, then a monthly doubling time is what you'd get. Important em politics, communication and organization happens at x1000 speeds as well, or at whatever speed is economically feasible.

Eliezer Yudkowsky writes:

Bryan, what do you think you know about molecular nanotechnology that Eric Drexler doesn't know and that makes molecular nanotechnology impossible? I can put no other interpretation on your claim that an em economy can't double in a year.

Or for that matter, sufficiently advanced protein biotechnology though God knows why they'd bother. Just what is this bottleneck in the infrastructure that no amount of technological advancement can double in a year?

Comments for this entry have been closed
Return to top