The era of emulation and what it means for us

Robin Hanson
Robin Hanson

What is the next phase for humanity? Robin Hanson set out to answer this question in a thought-provoking and lively talk to London Futurists on March 19th.

He argues that humanity has been through several distinct economic growth phases each of which has been “exponential” in character. The first lasted nearly 200,000 years from the moment Homo Sapiens first emerged as hunter-gatherers. These early humans were vastly superior to the animals they replaced, successfully exploiting their environment through the use of organisation and tools. The next economic era began with the arrival of agriculture about 10,000 years ago and brought about a huge acceleration in development, with efficient use of labour and larger and more sophisticated societies. This ended with the birth of the third era, the industrial era, which started around 1760. Again, an exponential increase in economic output and efficiency. This gave way to the computer age in which we currently are. The exponential periods of these eras has been becoming shorter and shorter with world GDP doubling roughly every 15 to 20 years today.

What, Hanson asked, could create an economy which doubles every week or month?

And the answer he comes up with is – robots.

What Hanson means by “robots” is true general artificial intelligence and he argues there are three ways to do this: better software, a comprehensive theory of intelligence, or emulating a human brain.

And it is his belief that the most likely scenario is that we will first develop the capacity to emulate a human brain and that this should happen “sometime in the next century”.

All we need, he argues, are “many parallel computers” which are capable of scanning a human brain, modelling every brain cell type and recording what we see and then “running the model”.

This doesn’t mean we need to understand how a brain works – he thinks we may be centuries away from this. But we would be able to run what he calls “EMs” – short for emulations.

If we had them there would be a new age – the age of EM.

It is this new era, then, that he sets out to describe. Running in software, EMs are effectively immortal -“like houses and cars, if we choose”. But it’s unlikely EMs will choose to be – much more likely that they will spawn short-lived versions of themselves to carry out repetitive or one-off tasks and then shut these down when they have served their purpose.

The new age will have new morals – EMs will probably be OK with termination and respooling.

Partly this is simply a result of obsolescence – “Currently if the economy doubles every 15 years your skills as an individual become obsolete in that time.” This is why we retire and let the next generation learn the next set of skills. “In the world of the EM faster emulation means faster obsolescence.”

They will run faster because, even though these new consciouses are essentially human brains, “human brains are parallel so more hardware means more speed.” And they will take up very little space as they only really need to inhabit robot bodies when they need to do something in the physical instead of the virtual world. Hanson believes most of the time they will inhabit a purely virtual environment.

Hanson sees the birth of EMs as inevitable – they will be developed to speed economic development. And in the early days humans will own the EMs – much like slaves were owned. But just like slaves, some EMs will “buy” their freedom and from there they will quickly make up more and more of the economy (which may now be doubling in a matter of weeks or days).  Because they are so cheap to create (an EM could be copied millions of times at very little cost) and because they cost so little to run he says wages will effectively fall to way below human subsistence wages.

Humans will be eclipsed. The whole human race will retire.

Whether that retirement is a happy or a tragic one is very much up to us, he believes, as we will be quite rich enough as a whole to ensure a good outcome, although those riches will be extremely unequally distributed.

But either way, we might be retiring into a very different world. “Robots don’t need nature” he says. “They may choose to save nature but don’t need to.”

And if we are thinking all this doesn’t sound too good, and that we humans are bound to resist, he doesn’t really buy the “robot wars” scenario, either. “There wasn’t a farmer-industry war during the switch to the industrial era.”

So if this new era could begin soon, how long will it last? Hanson believes that because EMs will be running so fast the whole era could last just a couple of years. After that, maybe they will develop true software AI which will spawn the next era – who knows….

Robin Hanson is an associate professor of economics at George Mason University and a research associate at the Future of Humanity Institute of Oxford University.

More promise for solar breakthroughs

Sample 3D Modules
Sample 3D Modules Credit: Allegra Boverman

The 3D solar towers being developed by MIT illustrate just how much further we have to go in stretching the still new technology of solar power. These experimental towers are up to 20 times as efficient as today’s flat roof panels.

If the British Government needed any further evidence that the plans for Hinkley Point are misguided, this should give them pause. Guaranteeing to pay three times today’s price for electricity for the next 30 years against a backdrop of dramatic increases in efficiency of solar power (which has already reached price parity in many places at today’s prices) is lunacy.

I would be tempted to conclude that the decision is ideological, but I just can’t see how a Conservative government getting into bed with a French state utility and the Chinese Government to make their white elephant a reality gels with the Conservative free market philosophy. Perhaps it is simply that having large grey boxes hidden away on the coasts are easier on the eye than rooftop solar panels and windfarms? Who knows.

Future Shock

There was a very perceptive article in the current issue of the Economist which argued, basically, that Moore’s Law is in sight of breaking down. The result, though, is maybe not what you might think. Progress may not necessarily just get slower; it is more likely to be much, much more unpredictable.

The reasons, according to the Economist are because these days there is so much more that is important than just the single chip in a single computer, among them the role of software, the cloud and new, specialised architectures optimised for particular tasks.

I think we can see some of this unpredictability unfolding in front of our eyes as Google’s Go-playing computer AlphaGo has beaten Lee Sedol, ranked number 4 in the world, in the first two of their best-of-five series. Go is seen as a special challenge to AI because it is very much more complex than chess and a “brute force” approach won’t work.

The really interesting thing about this match is that it was generally thought we were 10 years away from building a computer which could win at Go. AlphaGo surprised the world back in October last year when it won against Fan Hui who is ranked 633rd in the world. What has taken Lee Sedol by surprise is how much better the program has become since – we was apparently quite certain he could beat it.

Hold on to your seats – we could be in for some really quite startling surprises in the coming months and years.