Will IBM's Watson put your job in jeopardy?February 15, 2011: 9:33 AM ET
If IBM's supercomputer outperforms some of Jeopardy!'s greatest minds, what does that spell for the rest of us working stiffs? As it turns out, workers may have some cause for concern.
By Martin Ford, contributor
This week's showdown between IBM's Watson supercomputer and the world's top Jeopardy! players raises several questions on the impact that artificial intelligence will have on the future job market. After all, if a machine can beat humans at Jeopardy!, will computers soon be competing with people for knowledge-based jobs?
If IBM's (IBM) hopes for the technology are realized, workers may, in fact, have cause for concern. The company's website says that "Watson's first test will be on Jeopardy!, but the real test will be applying the underlying data management and analytics technology across different industries." In a video describing Watson, an IBM executive says the technology "will revolutionize industries at a level which has never been done before."
Unlike other advanced software applications, Watson has a command of natural language, it can simultaneously launch hundreds of information-seeking algorithms, and, perhaps most importantly, it has the ability to learn and improve its performance over time.
Watson's creators trained it with thousands of Jeopardy! sample questions, eventually driving it to perform at championship level. Just as practice makes a person better at a particular task, machine learning techniques allow software applications to get better with experience. Watson-like technology has obvious applications in areas like customer service and support, legal research and medical diagnosis. And when it comes to learning, machines enjoy an important advantage over human workers: once a machine or software application has been trained, that knowledge and expertise can be replicated quickly and easily.
Current research in artificial intelligence goes far beyond simply building machines to answer questions based on existing information. At Cornell University, a team of researchers has built an application that analyzes raw scientific data and is then able to discover new rules or equations explaining the underlying phenomena -- an accomplishment that might require years for a human scientist. In the future, it's likely that smart applications will increasingly combine natural language capability with the ability to autonomously find, analyze and present information in just about any field of expertise.
Watson is by no means a true thinking machine. Rather, it's a highly specialized application designed especially to play Jeopardy!. All existing practical applications of artificial intelligence are similarly specialized -- we are at least decades away from creating general, human-like artificial intelligence, and it may even be unachievable.
But don't assume that means artificial intelligence won't replace workers. Nearly all jobs in today's economy are specialized, and as applications like Watson become more versatile and affordable, they will be used in a variety of areas, especially in large organizations.
Machine learning is not limited to knowledge-based jobs and tasks. Heartland Robotics, a startup company founded by Rodney Brooks, a researcher at MIT, is reportedly developing a trainable manufacturing robot that will cost as little as $5,000 -- a price point that makes up less than two months of a typical worker's pay and benefits.
While Heartland is initially focusing on manufacturing, the most significant impact on workers in the United States will come if and when low-cost, trainable robots and other forms of automation are used in the service sector.
For many lower wage jobs, automation has been kept at bay by a human worker's unique ability to recognize complex visual images and then interact with their environment accordingly. But machines and robots are growing increasingly dexterous and better at seeing and understanding the world around them. Robots in Japan, for example, are able to autonomously pick strawberries, selecting only the ripest berries based on their color.
The technologies that power Watson will likely find their way into a variety of software applications and robots that can compete for both high and low skill jobs. As artificial intelligence software improves and hardware becomes dramatically faster and more affordable over the coming decade, job creation in both low and high skill occupations risks falling short of expectations. And employers in a wide range of industries may increasingly choose technology over people. Few, if any, economists seem willing to acknowledge that scenario, but if it does come to pass, what we consider unacceptable levels of unemployment today could become the new normal tomorrow.
Martin Ford is the author of The Lights in the Tunnel: Automation, Accelerating Technology and the Economy of the Future and blogs regularly at http://econfuture.wordpress.com.
Coming soon: You Can't Fire Everyone
Have you been fired -- or been the one swinging the axe? Tell us your stories. We'll highlight the most interesting and instructional stories in a section debuting in March.