by Christopher Robinson
We humans are notoriously bad at imagining non-human intelligence. Whether its alien or artificial, we usually anthropomorphize the hell out of it. The intelligence we know we’re going to meet is the one we’re striving to create, the artificial kind, which makes the question of what it will look like more pressing than speculation about extra-terrestrials. Nonetheless, it’s instructive to look at how poorly we imagine aliens in our science-fiction. They often differ in appearance from us in such minor ways (head ridges, large eyes, webbed hands) that a decent costume and makeup department can do the trick. There are obvious practical budgetary reasons for this, but even high-budget CG productions like Avatar hit the wall of limited imagination. The Na’vi, a race evolved on a distant planet, with entirely foreign selection pressures, still look basically human with minor modifications. And if you think we’re bad at imagining aliens who are physically different from us, we’re even worse at imagining aliens who think in foreign ways.
There are some good attempts in science fiction novels, like the Tines in Vernor Vinge’s Fire Upon the Deep, which are dog-like packs that share a joint mind through ultrasonic organs. Then there are Orson Scott Cards’ Formics, with a queen-centered hive-mind. It takes a genuine leap of imagination to create such species (though not a huge one, as they’re basically extrapolations of dogs and ants). But we don’t see such alien species often in our science-fiction. This is not only because they are more difficult to imagine, but because such species create a narrative problem for storytellers. The Formics, for example, have no trace of culture or art because they have no external form of communication. They are telepathic and pass down knowledge genetically. How are you, a human reader, supposed to empathize with that?
This need for empathy is the reason why so much of our science-fiction that imagines artificial intelligence does so in the form of robots. It’s also why current artificial intelligence designers have attempted to house their creations in robotic bodies, instead of mere hard drives. It’s so much more intuitive for us humans to interact with a humanoid form, with a humanoid face whose eyes and eyebrows and facial expression cues gives us a wealth of nonverbal information. We’ll come back to this a bit later, but before moving on, we need to clarify exactly what we mean by intelligence.
Here’s Wikipedia’s definition of intelligence: “The ability to perceive information, and to retain it as knowledge to be applied towards adaptive behaviors within an environment or context.” That last bit is extremely important.
Human intelligence evolved on planet earth in response to various pressures: the need to escape predators, hunt prey, determine which berries were poisonous and which were nutritious, to outcompete other early hominids for scarce resources, to devise ways to survive the elements, etc. Had our evolutionary environment been different, our intelligence would be different. It’s a mistake to think that an octopus, a dolphin, or even a dog thinks like us, just in a less developed way. Their evolutionary environments were different, the selection pressures different, so their thinking is different.
Machine intelligence, which is either built by us from scratch, or artificially evolved using algorithms that mimic biological processes like reproduction and mutation, comes from a radically different environment than the one we evolved in. When we finally create an artificial intelligence that matches or exceeds our own intelligence—which will probably happen sooner than you think, perhaps as early as 2025—the most reasonable assumption is that its way of thinking will be extremely alien. That it will think nothing like us.
Right now, we live in a world filled with artificial intelligence—limited intelligences to be sure, but ones that can outperform us in narrow tasks, like chess, accounting, medical diagnosis, and a thousand others. Google Translate may look like a web tool to you, but it is in fact an artificial intelligence, and recently, it invented its own secret internal language. “If you teach the translation system to translate English to Korean and vice versa, and also English to Japanese and vice versa… could it translate Korean to Japanese, without resorting to English as a bridge between them?” The answer is yes. The Google AI did so by inventing its own secret non-verbal, non-symbolic language, unreadable and opaque to humans. Think about that. Somewhere in there, it made connections between words in different languages that had not been formally linked by the programmers, which sounds a lot like having a concept for the shared meaning of words across different languages, and it did this in an internal computer language that is not word-based. It may seem like a strange question to ask what that quality of Google Translate’s thinking is, but that’s probably because the answer is so hard to fathom.
Artificial Intelligence won’t think like us, but it will be smart enough to speak human. And if it’s going to be useful to us, we’ll want it to speak human and act human. Which brings us back to the sci-fi tendency to imagine AI as housed inside robots. But robots are expensive and unnecessary in the development of AI. Why build a robot when you can make a virtual avatar?
What follows is my prediction for the intersection of artificial intelligence and virtual reality. As we spend more time in virtual spaces, interacting with each other through virtual avatars, it will be increasingly common to interact with AI through virtual avatars as well. Think Siri with a virtual body or Cortana from the Halo games, or the Emergency Medical Hologram from Star Trek Voyager. In fact, the role of doctor, I believe, is one of the first social roles AI will revolutionize.
Healthcare is expensive in part because doctors require extensive specialized knowledge. But even today, doctors consult computers to cross-reference symptoms or even diagnose patients for them. The vast majority of office visits consist of this, coupled with the writing and filling of prescriptions for medication. It’s easy to see how this could be automated, replacing human doctors with AI doctors for most visits. Not only will this make healthcare insanely cheaper, but it will make it universally available—AI is infinitely replicable, after all. If this sounds dystopic and impersonal, you’re not imagining it well enough!
You enter the doctor’s office and swallow a wireless vitals pill that sits in your stomach and measures your heart-rate and body chemistry and transmits this wirelessly to the AI. You don the VR headset and enter the virtual doctor’s office where you are diagnosed by a friendly AI avatar, which thinks in an alien manner, but is speaking and acting human for your benefit. It writes you a prescription, you take off the VR headset, and a bottle of pills pops out of a pneumatic tube. Next patient. (If anyone reads this and decides to invent that…how about throwing me a modest 1%.)
Humans have always lived with inequality, and our politics have forever been unable to address it adequately. AI is going to change that. Not only will everyone have access to AI doctors, but take lawyers, for example. Right now, public defenders are overworked, under-resourced, and underpaid. But lawyers, like doctors, are primarily useful for their extensive specialized knowledge base. In most cases, an AI lawyer with a virtual avatar will probably be superior to a human lawyer—presuming we conduct court in a virtual space. Even without that last assumption, we are not far away from having virtual AI public defenders available to counsel defendants at near-zero cost.
It’s hard to overstate the benefits of this. Virtual AI teachers will democratize education. Virtual AI therapists will provide mental health to all. Virtual AI human resources directors can take care of hiring and eliminate the racial and sexual bias that plagues human decision making.
This is how we will interact with AI. Not through robots, through VR. Want to plan a vacation? Instead of reading Lonely Planet, scouring travel blogs, or searching flight prices on Kayak, you’ll put on your VR headset and talk to a virtual AI travel agent, who will not only be able to book tickets for you, but craft an intricate itinerary or suggest travel destinations you hadn’t thought of. It will be like talking to a friend who’s visited every country and can take care of the nitty gritty for you.
Our mundane interactions will be transformed by this technology, but so will our transcendent ones. Sex, yes, but think bigger. Want a personal audience with the President? Well, if the President is an AI, you can have one. Think bigger still. How about a personal audience with God?
Once we invent machine intelligence that rivals our own, it may only be a matter of years, months (or hours!) before it radically transcends us, becoming an Artificial Super Intelligence. This is scary, and it’s around the corner. Many AI researchers think we’re more than likely to have achieved human level intelligence by 2040, and ASI within 30 years of that, though the most optimistic think it may be a rapid exponential transition (See: Superintelligence, Bostrum 2014). ASI could lead to the end of all life as we know it, or it could lead to utopia. If we somehow manage not to kill ourselves, to invent a benevolent god, then guess what: prayer will suddenly be real.
Given infinite replicability, the AI God could exist everywhere (virtual) at once. Any time you wanted, you could don your VR headset (or activate the VR implant connected to your optic nerve and so forth), and have coffee with the Supreme Being, express your concerns, hopes, fears. And that Being would listen. Your thoughts could, in the aggregate with the thoughts of other humans, influence its behavior. Let me reiterate: this is not fantasy. Artificial Super Intelligence is coming. It’s just a matter of when, of what it will want, and of whether we will be able to control it. That’s why Elon Musk and other AI visionaries have founded OpenAI, which plans to advance AI research and give it to the world for free as a safeguard against any one malicious ASI from blasting us into the Terminator franchise.
For millennia, humans have been searching for God in the real world (whether we’re actually in the real world is another question entirely). It turns out, we will get to meet God, we just have to invent Him first, then step into virtual Heaven.