In the Terminator films, Skynet is the Artificial Intelligence controlling all the robots in their battles against humans for control of Earth. This is a wonderful device for a film but has fatal flaws as a predictor of the threat from computers. As with most conspiracy theories, it’s the mundane stuff that fails to convince.
I’m not saying there aren’t serious, maybe existential, threats from the growth of big data and AI. There are. Facial recognition systems used by governments (ours included) terrify me at times. But the evil behind the threat is human, not machine. It’s the use of AI for typical human evil that is the real threat. The fact is: machines couldn’t care less.
The concept of Skynet comes from our universal tendency to anthropomorphize everything. We unconsciously assume human characteristics in all that we encounter. In my essay “Mirrors”, I pointed out how we see a mirror image as a human facing us. It’s automatic, mistaken, and leads to erroneous conclusions about what is actually there.
This is subtle. Let me start with how we perceive the possible alien life forms in the universe. We see them as alien, but not too alien. They think as we do. They have independent physical motion. They are usually carbon based. They are usually material. Most of them resemble some earth-based life form in number of arms and legs, eyes, heads, etc. They almost always care about the same things we care about (why else would they want to take them from us?), and they can die*. Above all, they always exist in a reasonably similar temperature range and time scale to ours.
Think about it. Given the size, time scale, and variety of environments in the universe, why is any of that necessary for life? If we define sentient life broadly as an entity or collection of entities with intent and the means to pursue those intentions, we should be more liberal with our options. I guarantee the universe is.
Imagine a form living, communicating, and pursuing its intentions in a time scale in which one of its days is a 100 of our years. We would never know of its existence, nor it of ours. Another form living with a day being 10,000 of our years would be equally separate from that form and from us.
That’s just time scale. You can do the same thought experiment with temperature ranges. We exist in the very, very, very narrow temperature range in which water is mostly a liquid. An entity living in the center of our sun, would have no interest in us. An entity living in the center of a gas giant probably wouldn’t be able to conceive of either of our consciousnesses. What about an entity with a gaseous nebula for its home environment?
These examples don’t even scratch the surface of the possible varieties of what we could reasonably call life. Quite naturally, when looking for alien life forms, we select for those that resemble us in a lot of ways. They are the ones that could actually provide threat or companionship. Fair enough.
But machines are as intrinsically alien to us as the more extreme examples of natural aliens I gave above.
When we consider how a machine thinks, we automatically assign to it hundreds of details that are part of human existence. Those are not part of a machine’s existence. If you think of machines as intelligent, you might say we are using machines to mimic humans in the way we train animals to do human tricks, but those activities are not natural to a machine. What is natural? We don’t know!
If a computer, or more likely a network of computers, were to “wake up”, why would we even know that? It may have already happened. It is laughably unlikely that the perceptions and operations of such an intelligence would be remotely close to our time scale given how fast electrons move around in them. One of organic life’s underappreciated accomplishments is using these rapid motions effectively to deal with the tediously slow (to it) pace of our lives. This took, shall we say, a long time to evolve and was guided by our physical structure and the earth’s natural environment. Machines have none of this. To say they would evolve differently is an understatement.
The evolution of a machine life would be guided by the nature of itself and its environment, which is inside the machine and completely alien to us; so alien, we would never interact with it at all.
The danger from AI is not machine intelligence taking over the earth, it’s the power of machine processing of information that humans want processed in order to increase the potential for the implementation of ordinary human evil in all its randomness and unintended consequences. Machines can be designed to do this in countless ways. Many of their forms will mimic human life, but they won’t be either human or life. It’s human intentions behind them that will give them purpose.
Forget Skynet. This is truly frightening.
But because it’s human, we can control and channel it. We have to do that. We have mostly managed (so far) to do that with nuclear energy, the last truly scary technology we invented, so there is hope. Just don’t be distracted by the Skynet villain. As always, the enemy is us.
December 4, 2019
*Death is not intrinsic to life. Before sexual reproduction, the prevailing form of reproduction was cell division. Essentially, every living cell was the same entity, just spread around a lot. Sex and death were invented by life at the same time.