These days, people worry about robots stealing our jobs. But maybe we should be more concerned about massive populations of computerized human brains. Called "ems," these infinitely-reproducible brains could change the world. Here's what will happen when digital brains vastly outnumber biological ones.
To learn more about this prospect, I spoke to economist and futurist Robin Hanson. He's an Associate Professor at George Mason University, a Research Associate at Oxford University's Future of Humanity Institute, and the Chief Scientist at Consensus Point. Hanson is currently writing a book on the economic and social implications of whole brain emulations — or what he simply refers to as "ems."
A brain emulation can be thought of as a type of brain upload. To make an em, a future neuroscientist would scan a particular human brain in high spatial and chemical resolution. Then, for all brain cell types, math models would be developed that map input signals to output signals and internal state changes. The final step would be to combine these to fill out a computer model of that particular brain.
"If the scan and cell models are good enough, the whole model must have the same input-output behavior as the original brain," Hanson told io9. "So if you add artificial eyes, ears, hands, and so on, it could talk with you and do tasks as well as the original. It could also do as well at arguing that it's conscious and deserves moral consideration."
"Ems would remake the world," says Hanson. "We humans are made of meat, our brains run at the same speed, we take decades to build, and we must be trained individually."
In contrast, argues Hanson, em brain models could be run at many different speeds in order to match different tasks. Being made of electronics, and built quickly in factories, the em population could grow very fast, allowing the em economy to also grow very fast. It could even double monthly.
"Because ems are easily copied, you could train one to be a good lawyer and then make a billion copies who are all good lawyers," he says. "That one initial em could come from the very best suited human; the typical em would be as sharp and capable as the very best humans in the world. The em economy would thus be much more competitive because small efficiency gains would lead to a bigger displacements of behavior."
Eventually, ems will be made cheaply — and they'll displace humans in virtually every niche. They'll basically do everything.
"Some would do physical jobs, like assembling or cleaning, and would have robotic bodies with sizes and shapes matched to their tasks," says Hanson. "But most would do desk jobs, like engineering or finance, while sitting in virtual reality offices. And almost all leisure time would be spent in virtual reality."
Since ems who are close enough can meet in virtual offices without having to move their brains, em cities have much lower travel congestion costs, allowing much bigger cities. "Picture a few huge tall hot cities packed dense with computer hardware, and with about half the volume devoted to pipes to move cooling fluids," says Hanson.
I asked Hanson why brain emulations would perform all these tasks, and not advanced algorithms.
"We have worked for centuries to automate jobs, and we've had a lot of success." he responded. Indeed, humans have been pushed out of many jobs. Even so, we still pay far more to "rent" human workers than to rent machines — showing that human workers still create far more value.
"I was a researcher in the field of artificial intelligence (AI) from thirty to twenty years ago, and I've asked lots of old AI hands how much progress they've seen in the last twenty years in the AI subfield they know best," he continued. "Their typical answer is that we've moved about five to ten percent of the distance to human level abilities, with no noticeable acceleration. At this rate it would take two to four centuries for machines to be as good as humans in most AI task areas. So we still have far to go."
In response to the question of why it's so hard to automate jobs, Hanson says that, because most brain activity is unconscious, we tend to vastly underestimate how much our minds do in simple-seeming tasks.
"It isn't just a few general algorithms; we have inherited a vast storehouse of tricks. There aren't five magic equations to make machines smart at everything, any more than there are for making successful cities or bacteria. As in biology or business, success requires not messing up on a million little niggly tasks."
Ems will (primarily) live in supercomputers and they'll be copied easily, so their populations will quickly skyrocket. This intense competition will have a profound impact on ems in terms of their market value as employees and as potentially rights-bearing citizens. I asked Hanson to elaborate on both these points.
"Supply and demand works well to analyze this," he says. "When we can make lots of em computers fast, then the wage for each type of job must fall to the cost of renting and running enough computer hardware for an em brain, plus the cost to train an em for that type of job divided by the number of trained copies, plus whatever extra payoff it takes to get two productive ems to compete for that job type."
In such a work environment, and with very cheap computer hardware, wages would be very low — near an em 'subsistence level,' and far less than what humans need to survive.
Hason agrees that minimum wage laws could force higher wages. But he says that such wages could put the jurisdictions that enact them at a competitive disadvantage in a very competitive world. And the higher that minimum wages rise above natural wages, the more eagerly business-folk would evade it via hidden copies of themselves. Draconian enforcement, warns Hanson, might be required.
"Now supply and demand doesn't tell you who owns stuff, only the price of trades in that stuff. So it doesn't tell you if ems would be slaves or fully autonomous citizens. But as we've seen in history, slavery offers little profit when wages are near subsistence; it was higher wages that enticed slavers."
Needless to say, an economy over-saturated with highly productive and motivated digital brains will have a tremendous impact on biological humans. To cope, meat-based humans will have to find ways to add meaning to their lives. Disturbingly, Hanson predicts that human servants could become status symbols among the rich.
"But to survive, most humans would need assets other than their ability to work," Hanson told io9. "Like stocks, bonds, real estate, or patents. Humans without such assets may or may not get charity from other humans. Em charity to humans may focus on offering humans a chance to become ems."
Disturbing, no? Hanson made it clear to me that his analysis is designed to forecast likely outcomes, and not those outcomes we currently desire.
"We first need a clear idea of what is likely if we do nothing, before we can decide where to push the future away from that no-pushing scenario," he says.
As noted, digital brains will function at different speeds, so some ems will be faster than others. Given the type of uber-competitive environment Hanson describes, this will likely create severe social and economic stratifications.
According to Hanson, it would cost twice as much for hardware to run an em mind twice as fast. So very cheap ems might run at one-thousandth of human speed, while ems costing a billion times more might run a million times faster than humans.
This could result in em subtypes, including kilo-ems and milli-ems, who think and execute tasks a thousand times faster or slower than humans, and potentially micro-ems, who run a million times faster or slower.
"Ems who manage small fast physical systems like nano-factories must be fast," says Hanson. "Those who run big slow systems like supertankers can be slow. Fast bosses could manage many diverse slower subordinates, enabling bigger organizations. Most ems would run near the speed of their interaction partners."
Over time, however, adaptive complex systems tend to get fragile (they become worse at adapting to new contexts). Brains, says Hanson, are no exception. This will likely put a limit of perhaps several centuries on the useful subjective career of ems. So, to fit this career into an economic doubling time, the typical em office worker may run a thousand times faster than humans.
"Ems running at different speeds would find it hard to interact naturally, and their cultures would quickly get out of synch," predicts Hanson. "Faster ems would have many markers of higher status, including more wealth. So yes, em society would stratify by speed."
Indeed, most ems would find it excruciating to converse with biological humans. From the perspective of a milli-em, a single conversation could require so much time it would not be worth the trouble.
Perhaps surprisingly, Hanson also predicts that some ems will voluntarily choose to have short lives.
"Compared to an em who works eight hours a day and then rests for 16, a copy of that em who works for eight hours and then ends would incur only one third the hardware costs per work hour," he says. "This is such a huge savings that most em workers would be such short-lived ems."
Alternatively, and to reduce stress, Hanson says it could be affordable to retire this copy to a slow life of leisure, instead of ending it (the slower the functional speed, the cheaper it will be to live). The main obstacles to an indefinite duration of retirement, however, would be low interest rates, or civilization-wide instabilities like revolutions or wars.
"In virtual reality, ems need never experience physical pain, hunger, disease, or grit," says Hanson. "Furthermore, spectacular quality virtualities would cost little compared to the cost of running an em brain. While retiree minds would eventually become fragile, frail minds can still enjoy life. And retirees might play key symbolic roles in em rituals."
Given these near-Dystopian scenarios, I asked Hanson what would motivate ems and why they would choose to cooperate with traditional power structures.
"Since ems are psychologically human, they would deal with hierarchy, group divisions, and coalition politics very much like humans do," he explained. "The main difference is that ems would have a new unit of organization: the clan of copies of the same original human. Since ems could trust their clans more, clans may become the strongest unit of social organization. Clans may negotiate and invest for their ems and offer them life coaching based on the experiences of millions of similar copies."
As if a lot of this isn't already grim enough, ems will have to deal with a host of other concerns.
For example, a stolen copy of an em mind could be tortured to extract secrets, or enslaved to compete with the original. Even a credible threat to steal and torture a copy could also confer great power.
"So most ems would be careful to avoid mind theft," he says. "Some, however, would offer themselves as open source ems, free to copy and ready to do any task assigned them."
Not-yet-retired ems would also be wary of entering unknown virtualities that might be so pleasurable or engrossing as to tempt them not to return to work.
Hanson assumes low regulation scenarios because he says low base scenarios are easier to analyze. But eventually, many jurisdictions would regulate many things
"And also because a more competitive em economy will select more strongly against jurisdictions whose regulations create competitive disadvantages," he adds. "For example, if most human-dominated jurisdictions are slow and cautious regarding the first ems, the em economy would blossom in the few places that allow quick adoption of em-friendly practices. Such places would soon dominate the world."
Hanson says the most likely regulations are those that will create competitive advantages. Those could include intellectual property rules that encourage innovation, and combinatorial auctions for the allocation, sizing, and pricing of urban land use and utilities.
Regardless, the future that Hanson describes is a far cry from what many of us are hoping for. Living as a stream of 1's and 0's may have its benefits — but a digital utopia has never seemed more elusive.