There’s a growing fear that artificial intelligence (A.I.) and machine learning will take over millions of human jobs, leading to social upheaval on an unimaginable scale. While some tech leaders believe that artificial-intelligence platforms will ultimately prove a blessing to the human race, there are others—most notably Tesla CEO Elon Musk—who think that unmanaged A.I. could doom us all.
At this juncture in the evolution of A.I., it’s too soon to tell how things will ultimately go. But there are a few signs that A.I. won’t undermine human beings, at least in the near term.
Exhibit A: the recent tournament that saw a bot created by OpenAI, a non-profit A.I. research company, beat human champions at “Dota 2,” a real-time strategy game. When news of the bot’s victory hit the Web last week, it was upheld as evidence that A.I. platforms are quickly learning how to dominate homo sapiens in arenas that require speedy thinking and improvisation.
And to a certain extent, that’s true: the bot really did adapt to the point where it could wipe out expert players. “In the span of a month, our system went from barely matching a high-ranked player to beating the top pros and has continued to improve since then,” read a new note on OpenAI’s blog. “Supervised deep learning systems can only be as good as their training datasets, but in self-play systems, the available data improves automatically as the agent gets better.”
But OpenAI’s bot—which relies on a combination of observations, actions, and feedback—had a very difficult time dealing with aggressive improvisation on the part of human competitors. “It can still be confused in situations very different from what it’s seen,” the blog posting continued. “We’ll need a system which can handle totally weird and wacky situations it’s never seen.”
Most human jobs have an element of “weird and wacky.” Take truck driving, which is a key focus of technologists working on autonomous-driving technology. Although the majority of truck driving proceeds within a real-world framework—the truck motors from Point A to B, obeying clearly defined laws and regulations that can be easily programmed into an autonomous system—there are definitely situations in which things go chaotic and weird: debris in the roadway, unexpected route closures, and so on. A human driver can adapt fairly quickly to these unexpected situations; but a robot, less so.
That same principle extends to programming and other, similarly sedentary professions: no matter how routine the day-to-day, new situations will crop up, potentially confusing bots and A.I. platforms to the point of inaction. (Automation has scaled up rapidly in a manufacturing context precisely because an assembly chain is such a predictable environment.)
Yes, A.I. systems will only get better at improvisation. Nonetheless, it might be some time before the software reaches the point where most companies are comfortable with handing over operations completely to a machine. Depending on your role within the tech industry, you may have quite a few years before you need to worry about the potential impact of A.I. on your employment.