It was oddly appropriate that director James Cameron introduced the world to Skynet—the fictional super AI which sought to eradicate humanity—in 1984.
According to Terminator lore, Skynet was created in the then-future 1990s to remove the human element from US nuclear defenses. But then Skynet became self-aware, initiated a global nuclear holocaust, and created an army of killer bots to take out the survivors, yadda yadda yadda.
Of course, this future dystopia was conceived long before anything like capable robots or artificial intelligence even existed. Fast forward to 2017 and human-optional tech isn't only out in the real world, but engineers are scrambling to devise ways to give them even more responsibilities. All around the world, autonomous mini-Skynets are becoming a (hopefully benevolent?) reality.
While we probably won't be handing something as precarious as the nuclear launch codes over to an algorithm anytime soon, society is growing increasingly reliant on technology to run other vital tasks. In fact, that world has become so complex that it's practically a necessity. Our infrasturcutre isn't just coming online, it's gaining the ability to anticipate and react. We've tasked our algorithms with spotting security breaches in complex systems, trading the majority of the world's stocks, and even predicting when things like plane engine parts might break before it happens.
To that end, engineers are increasingly utilizing things like "digital twins" to help make predictions and decisions. Digital twins are virtual representations of real objects (typically vital infrastructure like turbines in a power plant). These twins utilize real-time data to predict when something might fail (thereby allowing the upkeepers—which are themselves increasingly automated—to fix problems before they occur). But if AI is a type of intellect, would it be accurate to describe digital twins as a form of imagination?
"Yes, it is. But it's an imagination centered around what it actually knows and its past history, as well as about the environment and how you're using it," explains Dr. Colin Parris, VP of Software Research at General Electric and a leading developer of digital twin tech who was a recent guest on PCMag's interview series, The Convo. "That imagination tells it 'well based upon this data, I may need to be maintained at this time.'"
But digital twins aren't relegated to input from a single source—they are able to utilize the experiences of a whole fleet. If the algorithm, for example, observes that a specific plane part begins to experience wear after 2,000 landings in rainy conditions, then it can ping the upkeep crews the next time the plane goes in for servicing. But giving a system true intelligence is more than the "time for a check-up" light on your car's dashboard; it's about improving its capability over time.
A field of AI called "machine learning" allows computers to master tasks independent of human guidence. This stitching together of collected experiences facilitates a hive-mind that makes up for a lack of common sense. Without this digital zeitgeist, complex technologies such as self-driving cars would never be possible.
A single human programmer—or even an army of programmers—could never craft software to anticipate every real-world road scenario, but self-driving cars can learn by observation. For example, a self-driving car might not recognize a person in a wheelchair, but by observing how humans react to this novel shape that shares features with a person and a car, the software can learn that this is a kind of pedestrian who should be treated as such.
Not only does the software improve by watching human drivers' behaviors, it also records what has worked when other self-driving cars were on the road (and perhaps more importantly, what didn't). This communal learning allows machines to navigate a complex world with many unforeseen variables.
When you combine virtual modeling and predictive technologies with advancements in robotics, you can see how infrastructure will become even more autonomous moving forward. This automation is problematic from an unemployment view, but isn't necessarily a complete loss for humanity.
"There are some jobs that are dull, dirty, and dangerous. I want to make sure we don't have humans too often in those jobs," Parris explains. "I'll give you an example. We have oil rigs out in the middle of the ocean which have giant stacks that they use to burn off fuel. Somebody has to go up those stacks and see if it has rust on it—that's 200 feet in the air, they're hanging by a rope, there are gale-force winds up there. The chances of a mistake are huge. But now we have drones. The drones fly up there and fly in a circle and take pictures. The software analyzes where the rust and damage is. So now we don't have to put humans in a dangerous place."
As robots become tinier, smarter, and more capable, you can see how the systems that civilization depends on might learn to maintain (and possibly even repair and build) themselves. It's almost as if they are evolving into life-like systems, which can learn, imagine, and anticipate. Hopefully they won't decide to destroy us one day.
The Convo is PCMag's interview series hosted by features editor Evan Dashevsky (@haldash). Each episode is broadcast live on PCMag's Facebook page, where viewers are invited to ask guests questions in the comments. Each episode is then made available on our YouTube page and available for free as an audio podcast, which you can subscribe to on iTunes or on the podcast platform of your choice.