Robin Hanson updated analysis of the economic impact of Whole Brain Emulation Artificial Intelligence

Robin Hanson’s best guess for the next revolution on the scale of the industrial, or farming, or human revolutions, is artificial intelligence in the form of whole brain emulations, or “ems.”

The powerpoint slides for the talk are here

Sander Olson interviewed Robin Hanson in 2010

You have written about how a major AI advance could bring about a step change in economic growth. Could another breakthrough, such as fusion power or molecular manufacturing, lead to similar growth?

Answer: We actually spend less than 10% of the economy on energy, so it is difficult to see how an energy advance by itself could bring about exponential changes in economic growth. Molecular manufacturing is more plausible, but it is still a long shot. We only spend about 15% of the economy on manufacturing, though advanced nanotech could facilitate a cascade of changes. So artificial intelligence is the most plausible way towards much faster economic growth rates in a short timeframe.

An earlier version of the talk was given at the Foresight 2010 conference.

Robin Hanson: “Economics of Nanotech and AI” at Foresight 2010 Conference from Foresight Institute on Vimeo.

Cheap for Robots

* Immortality – yet maybe can’t afford
* Travel – transmit to new body (but security?)
* Nature – don’t need ecosystems
* Labor – work is less tool/capital intensive
* Copies!
* Malthusian population explosion, rapid growth
– Wages may fall to fast-falling hardware cost
– Depends on mental-task landscape shape
– Happens if they slave or if free
– Only Draconian population/wage laws could stop

Emulations Feel Human

* They remember a human life

* Retain human tendencies/personalities
love, gossip, argue, sing, play, laugh, …

* More alienated worlds (as were farms, factories)
– Office work, leisure in virtual reality
– Physical work (~20%?) in android bodies

* More unequal abilities, like our fantasy
– Can run minds faster, use wildly different bodies

* Loyalty testable via sims, shallow mind reading

* Many won’t believe are conscious, or “is me”
– But same social implications, few dozen is plenty

Humans Eclipsed

* Wages well below human subsistence
– Some humans may find servant jobs

* But rich if held non-wage assets
– Investments double as fast as economy

* Robot-Human war unlikely if integrated
– Live/work together, or use same law among/between
E.g., we don’t kill retirees, take their stuff

* Most emulations are of the few best humans
– Workaholic, ok with alienation & oft lose memories
– First ems win? Billionaires? No kids? Few fems?

More Implications

* Fast growth discourages transport, encourages local production

* Options: copies rent, loan, or slave
– Renter, borrow evicted if no pay?
– Min wage risks hidden slaves
* To recoup train, clan limits wage?
– Also regulate to keep clan reputation
– Most memory of long train, vacation?

* Security to prevent bootleg copies!
– Hold copies co-responsible re law?

* less than 1cm tall robots seem feasible
– Mind/body can speed up as size falls
– 7cm tall and greater than x365, subjective year in day

* One skyscraper holds billions – is megacity

Robin imagines zones of tinier and faster copies interacting in a more dense environment.

Combination of technology in the 10-25 year time frame could enable some of this

A precursor to this could be possible with the sky city factory mass produced skyscraper architecture of Broad Group combined with robotic avatars. DARPA and other are working on advanced teleoperation of robotic bodies combined with AI that would on the path to but inferior to whole brain emulation.

Sky City would be a 200 story building that could hold about 100,000 people and would cost about $1.25 billion.

DARPA has recently proposed a humanoid robot grand challenge.

By 2030, with advanced avatars and humanoid robots and early molecular nanotechnology, the robots would have zettaflop processing and some quantum computer co-processing. A 10 to 100 times multiplier effect in producivity could occur even before there is whole brain emulation.

The Skyscraper could hold 100,000 people and 1 million to 10 million robotic servants of different shapes and sizes.

Improvements in the factory mass production methods and processes (and the lowering of costs for the Nth copy) could lower the costs of the factory mass produced skyscraper by 2 to 3 times.

$500 million for each of 10,000 sky cities would be enough space for 1 billion people at a total cost of $5 trillion. $100 trillion will be spent on construction in the world in 1 decade.