Макс Тегмарк - Life 3.0 - Being Human in the Age of Artificial Intelligence

How Much Can You Compute?

After exploring how long future life can last, let’s explore how long it might want to last. Although you might find it natural to want to live as long as possible, Freeman Dyson also gave a more quantitative argument for this desire: the cost of computation drops when you compute slowly, so you’ll ultimately get more done if you slow things down as much as possible. Freeman even calculated that if our Universe keeps expanding and cooling forever, an infinite amount of computation might be possible.

Slow doesn’t necessarily mean boring: if future life lives in a simulated world, its subjectively experienced flow of time need not have anything to do with the glacial pace at which the simulation is being run in the outside world, so the prospects of infinite computation could translate into subjective immortality for simulated life forms. Cosmologist Frank Tipler has built on this idea to speculate that you could also achieve subjective immortality in the final moments before a Big Crunch by speeding up the computations toward infinity as the temperature and density skyrocketed.

Since dark energy appears to spoil both Freeman’s and Frank’s dreams of infinite computation, future superintelligence may prefer to burn through its energy supplies relatively quickly, to turn them into computations before running into problems such as cosmic horizons and proton decay. If maximizing total computation is the ultimate goal, the best strategy will be a trade-off between too slow (to avoid the aforementioned problems) and too fast (spending more energy than needed per computation).

Putting together everything we’ve explored in this chapter tells us that maximally efficient power plants and computers would enable superintelligent life to perform a mind-boggling amount of computation. Powering your thirteen-watt brain for a hundred years requires the energy in about half a milligram of matter—less than in a typical grain of sugar. Seth Lloyd’s work suggests that the brain could be made a quadrillion times more energy efficient, enabling that sugar grain to power a simulation of all human lives ever lived as well as thousands of times more people. If all the matter in our available Universe could be used to simulate people, that would enable over 10 69lives—or whatever else superintelligent AI preferred to do with its computational power. Even more lives would be possible if their simulations were run more slowly.9 Conversely, in his book Superintelligence, Nick Bostrom estimates that 10 58human lives could be simulated with more conservative assumptions about energy efficiency. However we slice and dice these numbers, they’re huge, as is our responsibility for ensuring that this future potential of life to flourish isn’t squandered. As Bostrom puts it: “If we represent all the happiness experienced during one entire such life by a single teardrop of joy, then the happiness of these souls could fill and refill the Earth’s oceans every second, and keep doing so for a hundred billion billion millennia. It is really important that we make sure these truly are tears of joy.”

Cosmic Hierarchies

The speed of light limits not only the spread of life, but also the nature of life, placing strong constraints on communication, consciousness and control. So if much of our cosmos eventually comes alive, what will this life be like?

Thought Hierarchies

Have you ever tried and failed to swat a fly with your hand? The reason that it can react faster than you is that it’s smaller, so that it takes less time for information to travel between its eyes, brain and muscles. This “bigger = slower” principle applies not only to biology, where the speed limit is set by how fast electrical signals can travel through neurons, but also to future cosmic life if no information can travel faster than light. So for an intelligent information-processing system, going big is a mixed blessing involving an interesting trade-off. On one hand, going bigger lets it contain more particles, which enable more complex thoughts. On the other hand, this slows down the rate at which it can have truly global thoughts, since it now takes longer for the relevant information to propagate to all its parts.

So if life engulfs our cosmos, what form will it choose: simple and fast, or complex and slow? I predict that it will make the same choice as Earth life has made: both! The denizens of Earth’s biosphere span a staggering range of sizes, from gargantuan two-hundred-ton blue whales down to the petite 10 -16kg bacterium Pelagibacter, believed to account for more biomass than all the world’s fish combined. Moreover, organisms that are large, complex and slow often mitigate their sluggishness by containing smaller modules that are simple and fast. For example, your blink reflex is extremely fast precisely because it’s implemented by a small and simple circuit that doesn’t involve most of your brain: if that hard-to-swat fly accidentally heads toward your eye, you’ll blink within a tenth of a second, long before the relevant information has had time to spread throughout your brain and make you consciously aware of what happened. By organizing its information processing into a hierarchy of modules, our biosphere manages to both have the cake and eat it, attaining both speed and complexity. We humans already use this same hierarchical strategy to optimize parallel computing.

Because internal communication is slow and costly, I expect advanced future cosmic life to do the same, so that computations will be done as locally as possible. If a computation is simple enough to do with a 1 kg computer, it’s counterproductive to spread it out over a galaxy-sized computer, since waiting for the information to be shared at the speed of light after each computational step causes a ridiculous delay of about 100,000 years per step.

What, if any, of this future information processing will be conscious in the sense of involving a subjective experience is a controversial and fascinating topic which we’ll explore in chapter 8. If consciousness requires the different parts of the system to be able to communicate with one another, then the thoughts of larger systems are by necessity slower. Whereas you or a future Earth-sized supercomputer can have many thoughts per second, a galaxy-sized mind could have only one thought every hundred thousand years, and a cosmic mind a billion light-years in size would only have time to have about ten thoughts in total before dark energy fragmented it into disconnected parts. On the other hand, these few precious thoughts and accompanying experiences might be quite deep!

Control Hierarchies

If thought itself is organized in a hierarchy spanning a wide range of scales, then what about power? In chapter 4, we explored how intelligent entities naturally organize themselves into power hierarchies in Nash equilibrium, where any entity would be worse off if they altered their strategy. The better the communication and transportation technology gets, the larger these hierarchies can grow. If superintelligence one day expands to cosmic scales, what will its power hierarchy be like? Will it be freewheeling and decentralized or highly authoritarian? Will cooperation be based mainly on mutual benefit or on coercion and threats?

To shed light on these questions, let’s consider both the carrot and the stick: What incentives are there for collaboration on cosmic scales, and what threats might be used to enforce it?

Controlling with the Carrot

On Earth, trade has been a traditional driver of cooperation because the relative difficulty of producing things varies across the planet. If mining a kilogram of silver costs 300 times more than mining a kilogram of copper in one region, but only 100 times more in another, they’ll both come out ahead by trading 200 kg of copper against 1 kg of silver. If one region has much higher technology than another, both can similarly benefit from trading high-tech goods against raw materials.