Presenter: Robin Hanson, Author of The Age of Em: Work, Love and Life when Robots Rule the Earth
If you were to scan a human brain and run a model with the same connections on a superfast computer, would the resulting intelligence be more human or robotic in design? Robin Hanson, author of The Age of Em: Work, Love and Life when Robots Rule the Earth, argues that while these emulations may be be distinctly technological, their psychological and even emotional makeup would cause them to feel as human as the individuals they mirror. This implication of a potential future of human-technology hybrids raises several questions: First, will emulations behave more as humans or robots? And what will happen to humans in this emulation-filled future society? Even with these questions unanswered, many argue that we are undoubtedly heading towards a future of artificial intelligence, with one big exception to the mainstream conception of AI – namely, that artificial intelligence will not be new intelligence, but virtual versions of ourselves.
The history of civilization contains key insights into the future of artificial intelligence. Throughout human history, progress has been characterized by steady growth rates and sudden bursts of activity where growth rates will increase dramatically by a factor of 50. Foragers, for example, doubled in growth every quarter of a million years; then, farmers every 10,000 years; and now, industry doubles in growth every 15 years. If this pattern were to continue, straightforward projections show that within the next century, economic growth would begin to double every month.
Smart robots are our next Industrial Revolution. With increasingly complex and effective software, we are an estimated two to four centuries away from full artificial intelligence capabilities; however, the possibility of undiscovered theories of intelligence may provide a catalyst for even sooner full AI capabilities.
Transitioning from one era to the next has always been a challenge. During the forager’s transition into farmers, emphasis on natural instincts was replaced by ideals like progress and culture. Within the last few hundred years, however, the issues that compelled us to become farmers have died down, and we’ve begun to reclaim some of our forager ideals, namely increased emphasis on democracy and art, and decreased emphasis on religion and war.
Rather than a forager-like future, predictions show that emulations will behave more as farmers. Due to the social-centered construction of emulations, AI technologies will most likely behave more like farmers – with a class system, hierarchy, and order – due to their low subsistence level income and lack of basic natural instincts that drive humans.
Emulations are the most likely form of artificial intelligence. Because it is easier to copy and paste intelligence than to build it from the ground up, our nearest form of artificial intelligence is most likely ourselves.
Emulating an already existing intelligence is easier than building a new intelligence from the ground up. If you have an old computer’s software that you want to run on a new computer, for example, you can either attempt to write new software that mimics the old software, or simply write an emulation of the old software. With the second (and easier) option, the new computer appears to be the old computer to the new software, and the new computer can run this new software without having to understand its origins.
Artificial intelligence emulations require three technologies to become a possibility. First, computers need to become cheaper, faster and more parallel; second, technology needs to be thorough enough to scan an individual human’s brain to find spatial and chemical details and connections; and third, intricate models need to be created for each kind of cell in the human brain while also detailing how the brain takes input signals, changes internal state, and makes output signals. With good enough models and scans, software can be developed to mimic the same input/output behavior as a human brain. And once this software is cheap, the age of emulations will be upon us.
A world of emulations only requires a few brains. In order to create an intricate system of self-aware emulations, only 12 people would be needed out of the seven billion humans today for brain scans and models.
All emulations share universal traits. Although emulations may take many different shapes and forms, there are several universal traits that all emulations share:
Emulations are files. Because emulations are, in essence, software, they have all of the capabilities that software has, plus more. First, they are “immortal”; second, they can travel at the speed of light; and third, they can be easily copied.
Emulations will work and reside within virtual reality. Because emulations reside within the virtual realm, their deference towards objects in reality such as nature and travelling will be nearly nonexistent. Instead, emulations will be able to create limitless cities and landscapes within the virtual world, and alter these areas to their liking. They will also not have to experience things like hunger, pain, disease or dying.
Emulation population expands much faster than human population. Due to emulations’ ease of copying, population can expand fast, and this extreme growth rate can consequently create an extreme increase in economic growth rates. However, this will also eventually lead to the emulations experiencing Malthusian subsistence levels much lower than human subsistence levels, as well as the inability to retire until a much later age than humans.
Emulations are social units. Because of their inherent multiplicity, emulations will be able to congregate and socialize with millions of other emulations just like them. These social units can then be used for government, law, and finance, and also as a means of giving and receiving advice and knowledge from their counterparts.
Emulations may begin as humans, but their software will enable them to become so much more. While an emulation’s style and characteristics will remain within the human range, these emulations will be based on our more productive counterparts – billionaires, gold medalists, Nobel Prize winners, etc. – in order to create as efficient an emulation population as possible. These emulations will be made up of human traits such as intelligence, conscientiousness, agreeableness, cooperativeness, and extraversion, which are some of the core traits that comprise the most successful humans today.
The way an emulation views life will be different than how we view life. Unlike humans, emulations have the ability to create copies of themselves to accomplish both temporary and long-term tasks, a trait that will undoubtedly help them both in work and life.
Emulations can work in both serial and parallel. Whether it be several copies at once to carry out immediate tasks or one copy at a time to look over a long-term project, emulations can use their copies in a variety of ways. While master planner emulations may oversee a whole system of their own copies and sub-copies working together on a large task, general worker emulations will create thousands of identical copies of themselves in order to accomplish different menial tasks.
Emulations will retire differently than humans. Because emulations are based on humans whose ability to adapt wanes with age, emulations will also most likely become specialized in only one or two areas. And although they have an infinite lifespan, their limited career life – which will span a mere one or two centuries – will end in indefinite retirement.
An emulation’s life plan is not linear. While humans generally follow a pattern of education, work, and retirement, an emulation’s ability to copy itself alters this typical life plan. For example, an emulation may split into three copies who are then educated by the original emulation, do three times the work, and then retire as three copies at a much faster rate than the original, who continues to work and create copies who then work and create copies in an endless cycle.
Emulations can run at different speeds. Although emulations can be run at speeds up to one million times as fast as a human or a billion times slower, their speed comes at a price.
The faster you run an emulation, the more it will cost you. While faster emulations will get more done faster, the cost for the emulation will consequently be much higher than an emulation running at normal speed. If an emulation retires poor, for example, they will need to slow down their speed to a speed they can afford.
Emulations will gather with other emulations running at a similar speed. In order to work at with others running at a similar speed, emulations will begin to clump into a small number of extremely dense virtual cities – something that isn’t possible with humans in the real world.
Differences in speed will naturally create a hierarchy. Because emulations running at a higher speed will acquire more knowledge, money, and status faster, a discrete social hierarchy will be created with speedy emulations at the top and slow retired emulations at the bottom.
A month in our world will be a century to an emulation. Because of their inherent speed, virtual reality will progress at a much faster rate than we can experience as humans. After a month in real time, a century will have passed in the virtual world and the economy will have doubled in both arenas.
To read or download the rest of the essays from this special report on the Future of Work and Education, download our free app on your favorite device (iStore, Google Play, and Amazon Kindle) or click to view the Digital Edition.