Confronting the challenges presented by the development of artificial intelligence
By M.S.
KEVIN DRUM looks to have set the topic for the day with his article in Mother Jones on the economics of our robotic future. The argument is a good recap of several points that have also turned up in speculation by others, including Paul Krugman and my colleague, on what happens once artificial intelligence surpasses human intelligence and robots start replacing us not just in manual labour or repetitive processing tasks, but in, well, everything. Assuming Moore's Law keeps churning away at its normal exponential pace, Mr Drum figures that will happen somewhere around 2040, and it will gradually make our current economic assumptions untenable: most humans will become permanently unemployable since there will be nothing they can do that a robot can't do better and cheaper, which means there will be too few consumers to create demand for the products the robots can create. The only way out will be to vest humans at birth with ownership shares in the robot means of production, as Noah Smith has suggested, creating a post-capitalist society of hereditary aristocratic humans and robot slaves. Karl Smith agrees, but adds a moral caveat:
What’s going to happen is massive income transfers to flesh and blood human beings. These income transfers will come to be seen as a right-of-birth. This will make complete social sense once you realize that most of the beings on earth will be robots and therefore not-of-birth.
Birth is something that happens to a minority of beings who are special, flesh and blood humans.
The concern, as I see it, is over accepting the dual truth that robots will in all likelihood be sentient beings with an inner life just as ourselves, and they will live in grinding inescapable poverty.
I think both Mr Drum and Mr Smith are failing to integrate one more special factor about the artificial-intelligence revolution, though. Here's the thing about robots: they will be telepaths. When we think about intelligent entities, we instinctively model them on our own experience, where the thoughts we have take place through lightning-fast interchanges between billions of neurons inside our brains, while connections to information sources outside our skulls take place via relatively slow, dumb, evocative means like language, vision and empathy. For robots it won't be like that: information processing via electromagnetic links with the cloud will be just another form of neural connection, much as my laptop right now is actually writing this blog post on a server thousands of miles away. In fact, it's quite likely that the first entity to achieve human-like levels of intelligence will be Google, rather than some metal humanoid. We're talking the Borg, not C3P0.
As the Borg example makes clear, telepathy takes the concept of individual identity and schmears it. From that perspective, I think Mr Smith's concerns about the injustice of treating humans as persons while conscious artificial intelligences are treated as slaves is insufficiently pessimistic. The real problem with AI's is that it won't even be clear where one AI stops and another one begins. If Google were a person, what would it encompass? Would it include my docs on Google Drive? I have a couple of Google tabs open right now. Who do those tabs belong to? Me? It? Its shareholders?
Even this, though, fails to take on what may be the real long-run challenge. Brain-computer interfaces are still in their infancy, but there's no reason to suppose they won't progress just as rapidly as AI itself. By 2040, people may well be communicating directly with servers through chips implanted in their brains, which is to say they may be communicating telepathically with each other. If two servers at Google headquarters are both part of Google, what are two humans linked by a terabit-per-second direct neural link? Ten humans? A million?
I think the writer who's addressed these concerns most clearly is Charlie Stross. In "Glasshouse", for example, Stross makes it clear that in a post-Singularity society the key concern becomes the protection of individual identity, because infinite access to information tends to make everything bleed into everything else.
Our economy and our sociopolitical structure are systems of activity embedded in a physical platform. They both presume that individual persons encased in a single skull can be treated as independent actors and given rights of ownership and legal responsibilities. This hasn't always been the case; other and earlier social systems and economies often worked with families as the unit capable of owning or being held legally responsible, and our own particular current system posits a legal entity called the "corporation" which can also own things and be held responsible for actions even though it has no physical body. Our system of making individuals independent and responsible works because we train individuals to act independent and responsible. I think one of the real challenges as artificial intelligence develops is that we're going to have to look increasingly at how intelligence emerges in systems that have no clear boundaries and can't be delineated as separate persons, and our political system and our economic laws may come to seem increasingly antiquated and baroque, designed for beings that no longer exist.
