Since 1987 - Covering the Fastest Computers in the World and the People Who Run Them

February 10, 2011

The Curse of Smarter Machines

Michael Feldman

This upcoming Jeopardy showdown between IBM Watson and grand champions Ken Jennings and Brad Rutter should make for great TV, especially for those of us who love to see cutting-edge computers in action. But if the machine performs as promised, the spectacle will demonstrate an uncomfortable truth: as machines get more adept at human-like calculation, even highly-skilled experts could become redundant.

With the ability to understand language, learn from experience and perform sophisticated analytics, Watson aims particularly high in this regard. Although the Jeopardy contest is just a PR demonstration, IBM has its eye on moving the technology into the commercial realm for things like business intelligence, financial analytics, medical diagnostics, and a host of other lucrative applications.

The human counterparts for these types of jobs tend to be well-educated and well-compensated individuals. And these “smart” machines are not all that expensive. A computer like Watson would probably cost in the neighborhood of $5 million today. That’s hardware only — we can only imagine what IBM would charge for the software, which is the real value add here. Even at $5M-plus, being able to replace one or more five-figure salaries with a machine that can work 24/7 would still be tempting.

To be fair, the Watson technology is not at a point where it could actually take the place of a medical diagnostician or a stock portfolio manager. Rather it would act as a support tool that could greatly magnify the performance of those individuals. The idea would be for a single analyst to perform the work of a dozen.

IT writer Nicholas Carr has expounded on this subject at length in his books and online blog. His take is that the advance of information technology is displacing the modern workforce at a rapid rate, just as the industrial revolution did for manual labor in the 18th and 19th centuries. And as the machines become more sophisticated, ever more highly-skilled jobs are being threatened.

In a 2007 blog post Carr writes:

In the past information technology tended to reduce demand for low-skilled jobs but increase demand for higher-skilled specialists. Now, automation is moving up the skills ladder, as the Internet and sophisticated software combine to reduce the need for more categories of knowledge and creative workers. One has to wonder what new categories of employment will expand to absorb the losses.

In another post he zeros in on software:

[I]f you look at more recent trends, you see that software is becoming increasingly more adept at taking over work that has traditionally required relatively high skills – or even, in YouTube’s case, enabling the creation of sophisticated goods through the large-scale and automated harvesting of free labor. The next wave of “superstars” may be algorithms – and the small number of people that control them.

Carr isn’t the only one to notice this. He cites a number of economists who have hypothesized that tech advancements may be one of the primary causes of the concentration of wealth for top earners. Fed chair Ben Bernanke noted that new technologies tend to increase the productivity of highly-skilled workers, and thus their wages, compared to lower-skilled workers.

“Considerable evidence supports the view that worker skills and advanced technology are complementary,” says Bernanke. “For example, economists have found that industries and firms that spend more on research and development or invest more in information technologies hire relatively more high-skilled workers and spend a relatively larger share of their payrolls on them.”

That would suggest that we just need to develop an increasingly higher-skilled workforce to keep pace with technological innovation. But a funny thing happens on the way up the food chain. The structure is really a pyramid, with fewer and fewer positions as you approach the top. For example, if you replace a maid with a robot, a single technician would be able to maintain many robots. So you just can’t retrain the maids to be technicians. The same would go for analysts as they get displaced by smart machines.

Today many economists are concerned about how slowly employment is recovering after the Great Recession. Sure enough, as the economic slide ended, productivity surged as companies discovered new ways to run their businesses with fewer people. I suspect a lot of that productivity surge was the result of more IT deployment rather than longer work hours. In many cases, businesses cut work hours to reduce costs.

So what happened to all the displaced workers from the recession? Well many are still looking for a path back into the workforce, while others have given up entirely.

Here’s an interesting graphic from the Bureau of Labor Statistics that shows the recovery of employment after the last six economic downturns:

Although the March 2010 New York Times article that cites this graph is making a point about the lag in re-employment, the more interesting fact is that the lag times appear to be lengthening significantly with each successive recession, regardless of the severity. That would suggest that job seekers are finding it increasingly difficult to return to work with each passing year. It’s not unreasonable to imagine that the inability of workers to keep pace with technology advancements is playing a role here.

If true, at some point that lag will be so long that employment won’t recover before the next recession hits. And then what? Well, we better hope that those new categories of employment Carr wonders about will actually come to pass.

Share This