Increasing Use of Autonomous Systems Could Threaten Jobs

Artificial intelligence experts predict that intelligent and semi-intelligent autonomous systems — such as self-driving cars and autonomous drones — "will march into our society" in the next two to three years, with driving expected to be fully automated in 25 years, a panel of experts said at a 13 February news briefing at the 2016 AAAS Annual Meeting.

"For the first time, we're going to see these machines and systems as part of our everyday life," said Bart Selman, professor of computer science at Cornell University, citing big changes in the AI field that have spurred a shift toward real-world applications in the last five years, including the ability of computers to see and hear as humans do and to synthesize data and fill in strategies for achieving their programmers' high-level goals.

Bart Selman | Ashley Gilleland / AAAS

With more than a billion dollars spent last year on AI research — more than in the field's entire history — the experts agreed that AI advances may threaten jobs and uncover a range of legal, regulatory, and ethical issues.

The widespread use of self-driving cars, for instance, is likely to bring about a reduction in car accidents; liability debates as courts determine whether a computer can be held at fault in an accident; and a serious effect on the labor market.

With 10% of U.S. jobs involving the operation of a vehicle, "we can expect the majority of these jobs will simply disappear," said Moshe Vardi, professor of computer science and director of the Ken Kennedy Institute for Information Technology at Rice University.

The expected disappearance of these jobs echoes trends in the manufacturing industry, Vardi said. U.S. manufacturing volume is currently at its peak, yet the number of U.S. manufacturing jobs peaked in 1980 and has now dropped below the 1950 numbers. Vardi attributed the falling number of jobs to automation.

Moshe Vardi | Ashley Gilleland / AAAS

"Today we have over 250,000 industrial robots in the United States, and the growth rate is in the double digits," said Vardi.

Vardi expects that the growing presence of intelligent machines in the workforce will contribute to a phenomenon called "job polarization." With many high-skilled jobs requiring too much human intelligence and many low-skilled jobs too expensive to automate, those jobs in the middle will be easiest to automate. The disappearance of these jobs will spur "great inequality," but even in a U.S. presidential election year, the issue is "nowhere on the radar screen," Vardi said.

Further decoupling of work and wages could radically restructure economic life in the long term, Vardi said. Since the agricultural revolution more than 10,000 years ago, human survival has been linked to work.

"We need to start thinking very seriously: What will humans do when machines can do almost everything?" Vardi said. "We have to redefine the meaning of good life without work."

Yet there are actions in the short term that policymakers can take to address potential ethical and legal concerns of AI machines such as autonomous weapons, said Wendell Wallach, an ethicist with Yale University's Interdisciplinary Center for Bioethics and the Hastings Center.

Wendell Wallach | Ashley Gilleland / AAAS

"There's a need for concerted action to keep technology a good servant and not let it become a dangerous master," Wallach said.

Wallach laid out several policy proposals to maximize the benefits and minimize the risks of AI machines, including the devotion of 10% of AI research funding to the study of the societal impact of AI machines. He also called for a presidential order to declare that, in the United States' view, lethal autonomous weapon systems violate existing international humanitarian law.

"We need strong, meaningful human control," Wallach said.