No they can't. Yes they can!
It's Man v. Machine on Jeopardy this week as IBM super-robot Watson
takes on former champions Ken Jennings and Brad Rutter. At The
Atlantic, we're using Watson as an occasion to think about what smart
robots mean for the American worker. This is Part One of a three-part
series on the exciting and sometimes scary capabilities of artificial
A common misconception about automation is the idea that it will primarily impact low paying jobs that require few skills or training. To illustrate that this is not necessarily the case, consider two very different occupations: a radiologist and a housekeeper.
A radiologist is a medical doctor who specializes in interpreting images generated by various medical scanning technologies. Before the advent of modern computer technology, radiologists focused exclusively on X-rays. This has now been expanded to include all types of medical imaging, including CT scans, PET scans, mammograms, etc. To become a radiologist you need to attend college for four years, and then medical school for another four. That is followed by another five years of internship and residency, and often even more specialized training after that. Radiology is one the most popular specialties for newly minted doctors because it offers relatively high pay and regular work hours; radiologists generally don't need to work weekends or handle emergencies.
In spite of the radiologist's training requirement of at least thirteen additional years beyond high school, it is conceptually quite easy to envision this job being automated. The primary focus of the job is to analyze and evaluate visual images. Furthermore, the parameters of each image are highly defined since they are often coming directly from a computerized scanning device.
Visual pattern recognition software is a rapidly developing field that has already produced significant results. The government currently has access to software that can help identify terrorists in airports based on visual analysis of security photographs. Real world tasks such as this are probably technically more difficult than analyzing a medical scan because the environment and objects in the image are far more varied.
Radiology is already subject to significant offshoring to India and other places. It is a simple matter to transmit digital scans to an overseas location for analysis. Indian doctors earn as little as 10 percent of what American radiologists are paid. As we saw earlier, automation will often come rapidly on the heels of offshoring, especially if the job focuses purely on technical analysis with little need for human interaction. Currently, U.S. demand for radiologists continues to expand because of the increase in use of diagnostic scans such as mammograms. However, this seems likely to slow as automation and offshoring advance and become bigger players in the future. The graduating medical students who are now rushing into radiology for its high pay and relative freedom from the annoyances of dealing with actual patients may eventually come to question the wisdom of their decision.
Is any job safe from automation?
Now let's turn to a very different job: that of a housekeeper. A housekeeper, of course, doesn't require any formal education at all, but as you might have guessed, this job is actually much harder to fully automate than the radiologist's. To take over the housekeeping job, we would need to build a very advanced robot--or perhaps several robots to perform various tasks.
If you asked the housekeeper to name the most difficult part of his or her job, you might expect the answer to be cleaning the bathrooms or the windows. For our robot, however, the truly difficult task is probably something that is relatively light work for the human housekeeper. Consider what is involved in tidying up clutter in a typical home. For the housekeeper, this is easy. A human being can instantly recognize objects that are out of place and can quickly put them back where they belong. Building a machine to reliably do the same thing is probably one of the most difficult challenges in robotics.