They are among us now.
Earth is the home of an unnatural intelligence unlike anything that has ever lived here. It is inhuman. It is intelligent. We might want to call it “alien,” but it was created on Earth, so our planet is its home. We call this intelligence “artificial” (and abbreviate it as “AI”) and, in this second decade of the 21st century, we are witnessing its rapid evolution and proliferation throughout the world.
Human intelligence is always personal. You and I view our lives from a unique perspective that is not replicable, the subjective perspective of a self. Machines do not yet have selves, but that will soon change. Human brains evolved with the ability to construct elaborate fictions that we call “personal identity,” and this capability is a feature of human intelligence that, over eons, survived the harsh competition of selective advantage. Why? Almost certainly, a personal identity assists its possessor in navigating a dangerous and unpredictable environment. It confers the ability to quickly integrate many disparate maps of reality, external and internal, into a single coherent representation that rapidly assists its owner in responding to ever-changing circumstances quickly and accurately. In other words, it assists us in being intelligent. If a personal identity is useful to intelligence, it will soon be coming to a machine near you.
Our ancestors have long found human personality traits in non-human beings, including other animals and features of the natural world such as mountains, rivers, and storms. And now we give personalities to machines. This tendency is so pervasive that we have a name for it: anthropomorphism. When we talk of “anthropomorphizing” a non-human entity, we express two distinct ideas: first, the object is being represented as having human-like features; and, equally importantly, that attribution is entirely unwarranted. When it comes to the AI invasion, this is the crucial consideration. These intelligences are not human. Their goals and concerns did not start to form in a human womb; their bodies did not emerge into this world during a human birth; and their personalities were not formed during a 20-year journey from infancy to adulthood.
When we examine the makeup of personal identity, we realize that it is deeply and necessarily biological. It certainly also contains features that are not biological (e.g. social, cultural, legal, semantic, etc.). It might even be possible in some imaginable future for people to completely abstract personal identity away from its biological roots. But, for now, human personal identity is bound to the brute facts of the human body. Pain, pleasure, hunger, fear, and love — the most elemental of all human experiences — are all based in our bodies. And we can’t really conceive of human selves that entirely lack the most fundamental elements of human experience. So human identity requires a human body. Another way of saying that is that no form of artificial intelligence has a human self. But that does not mean that artificial intelligence has no self at all; rather it means that if an artificially intelligent entity has a self, it cannot be human.
The problem with artificial selves does not arise simply because they are not human. It’s very likely that our world contains many selves that are not artificial and are also not human. Most people who interact frequently with dogs attribute them with selves, or canine identities. Certainly most primatologists would agree that mature chimpanzees have selves and even have a sense of personal identity that may be in many ways similar to that of humans. Humans have shared the planet with non-human selves for hundreds of thousands of years and we don’t think that such non-human selfhood is essentially problematic. The threat emerges when we combine artificiality with intelligence.
We don’t know that artificial selves are necessarily dangerous, but we do know that they are not human. The same thing could be said about fish selves (if we suppose they have selves at all). Generally that’s not a problem for us, but it might be if we are in the water with a fish and it happens to be a great white shark. And it certainly would be a problem if it turned out to be as dangerous as any shark, but was actually one of an exploding population of such creatures, each one much more intelligent than any human being. Especially so if they are not confined to the water.