Robots and AI aren’t coming for your job, just the boring parts of it
We extol the virtues of robots and AI quite a bit on this blog, and rightfully so; beyond serving as one of our competitive advantages in innovative solution design and development, it’s also one of the most exciting frontiers in modern computer science (which at this point is to say, the modern world). But as with any level of automation, one of the first questions to follow (again, rightfully so) is whether or not these advancements will mean an end to thousands or millions of jobs. The good news is that humans are still too good at too much for AI to replace many of the jobs it might have had in its crosshairs. The better news? Robots and AI together might be able to eliminate many of the more tedious, time-consuming or boring tasks away from our jobs.
Wouldn’t that be swell?
Robots and AI fall short alone
“We fleshy beings remain more creative, more dexterous, and more empathetic—a particularly important skill in health care and law enforcement,” according to Matt Simon over at Wired. “What is happening is that the machines are taking parts of jobs, which isn’t anything new in the history of human labor: Humans no longer harvest wheat by hand, but with combines; we no longer write everything by hand, but with highly efficient word processors.”
That sentiment holds true for many of the most crucial 21st-century jobs — it’s not about task completion so much as it is the soft skills to build relationships, collaborate effectively, unleash creativity and so on. And at this particular juncture, humans still have a pretty significant edge when it comes to those things.
Radiology proves the point
According to Erik Brynjolfsson, director of the MIT Initiative on the Digital Economy, radiology provides an excellent example of robots and AI automating parts of a job, without coming anywhere near replacing humans doing that work:
Let’s take one example. There are 27 distinct tasks that a radiologist does. One of them is reading medical images. A machine-learning algorithm might be 97 percent accurate, and a human might be 95 percent accurate, and you might think, OK, have the machine do it. Actually, that would be wrong. You’re better off having the machine do it and then have a human check it afterward. Then you go from 97 percent to 99 percent accuracy, because humans and machines make different kinds of mistakes.
But radiologists also consult with patients, coordinate care with other doctors, do all sorts of other things. Machine learning is pretty good at some