If there was one fear that dominated popular science in 2014, it was artificial intelligence. Stephen Hawking is scared it’ll kill us all. Elon Musk, the CEO of Tesla, called it “more dangerous than nukes.” But the truth is far different, and more complicated, than we think.
So what, precisely, is artificial intelligence?
That is an excellent question and once scientists are done arguing over the nature of intelligence and what quantifies it, we’ll have an answer. They’ve been at it for hundreds of years, so we might be waiting awhile.
More practically, there are a few schools of thought on artificial intelligence. Most commonly, though, the goal is not to create artificial life, but rather to create autonomous software to reach a goal set by a human. It’s generally called an “intelligent agent,” which admittedly is not the best terminology.
Don’t computers already do that?
Nope. Most computers follow a process that has been carefully laid out by a human being, and when you push a button, they go through that process step by step. This is most technology; your microwave won’t spontaneously make you mac and cheese, your GPS won’t grab your car and steer it to work, and so on.
Even most “dynamic” technology is really just following a more complicated process. When your GPS changes routes, it’s because it’s hit some pre-determined criteria to follow a new process. It seems intelligent, but it’s really just going through the motions. This is why your computer sometimes crashes; it can’t continue with a process and has no idea what to do.
How would artificial intelligence be any different?
An artificial intelligence would create its own process. Instead of a human telling it what to do, said human could just type in the goal and let the intelligent agent loose. This would save a lot of time, especially with robots, and generally create that push-button future the Jetsons promised us way back when. Or kill us all, depending on whom you believe.
So why does everybody think artificial intelligence will kill us all?
Um… there are a lot of movies about it? The main problem with AI panic is that it’s not based in any sort of reality.
Well, for example… why would an AI hate the human race? Why would an AI hate, or have any emotion, at all? We joke about robots feeling fear and shame, but those are either essentially complex toys or not emotions as such, but rather tools to teach the robot to not put itself in danger. We’re not talking “hate” as in “Kill all humans,” we’re talking about “hate” as in “I hate Nickelback.” Most people can’t stand said groaning Canadians, but if they caught somebody with a sniper rifle on a hill ready to pick them off, pretty much everyone would be horrified.
Which, in turn, raises the biggest issue with AI: We have no idea what the hell we’re talking about, at all, anywhere.
Wait, what? These are well-defined terms.
Not really. Like I said above, we’ve got no idea how intelligence or emotion work. For all our vaunted technology and medical advancements, nobody understands why humans laugh, for example. Nor is it really clear why we experience emotions in the first place, or how common emotions are in mammals.
Finally, all of this hinges on an artificial intelligence developing self-awareness, or “consciousness” for lack of a better term. But what’s “consciousness?” We have yet to quantify it scientifically. It’s still largely a philosophical discussion. How can we create something we quite literally don’t understand?
Are there no concerns about artificial intelligence?
No, there are plenty of reasons to be concerned. These are tools, and just like you can use a shovel to dig a ditch or beat someone senseless, these tools can be turned to malicious ends quite easily.
Just as an example, major corporations that know far more about us than we care to admit are developing more complex intelligent agents all the time, which can be used to predict future behavior. It’s uncomfortably easy to imagine a world where corporations, governments, and individuals use intelligent agents to model behavior and figure out everything from your medical history to your next career change. If that sounds paranoid, keep in mind that Target doesn’t even have this software yet and it’s shockingly good at doing just that.
Another problem might be that somebody could create an AI that, for example, just constantly does research. That sounds innocuous, but we probably wouldn’t understand the tools and products it would create beyond a certain point. It’d be like handing a wrench to a toddler. If you start imagining millions of intelligent agents doing this, it gets a bit scary rather quickly, especially if we’re talking about weapons.
So, forget the Terminator. In the end, the biggest threat to humanity when it comes to intelligent agents is likely our own greed and stupidity.