The current claims that automation, AI, and robots are producing widespread unemployment are preposterous. Martin Ford’s book “The Robots Are Coming” spread the fears raised by an Oxford University report that 47% of jobs were automatable by AI. These preposterous reports were clearly wrong, as unemployment has dropped to below 4%, even as automation, AI, and robots spread.
Automation eliminates certain jobs, but that disruptive process has been going on for hundreds of years, from at least the time when Gutenberg’s printing presses put scribes out of work. However, automation usually lowers costs and increases quality, leading to vastly expanded demand, which triggers expanded production to serve growing markets, bringing benefits to many people. The expanded production, broader distribution channels, and novel products lead to increased employment.
The exaggerations about dramatic technology revolutions goes back at least to 1970 when Alvin Toffler’s Future Shock declared that “What is occurring now is, in all likelihood, bigger, deeper, and more important than the industrial revolution… nothing less than the second great divide in human history, comparable in magnitude only with … the shift from barbarism to civilization.”
The lessons of history are familiar and clear.
- Weaving machines led to the Luddite revolt, but the dramatic drop in textile prices also led to much greater demand. The expanded production of cloth, the growing fashion industry, and new clothing distribution channels increased employment, while enabling more people to own more clothes.
- When Louis Daguerre announced photography in 1839, famed French artist Louis Delaroche declared that “from today painting is dead”, but the flourishing impressionist and other movements showed that there was much creative work to be done beyond photo-realistic landscapes and portraits. Photography, video, and visual communications expanded creative possibilities to become big businesses, enriching the lives of many people.
- In the 1960s, automated teller machines were supposed to put bank employees out of work, but the numbers of branches have increased as more people took advantage of expanded services such as mortgage loans and credit cards, thereby raising employment.
The main effect of automation is to lower costs and increase quality, which increases demand and expands production, bringing benefits to customers while increasing employment. Automation is also disruptive by making certain skills obsolete, so there are three challenges to technology innovators and policy makers:
(1) ENSURE SAFETY: design safe products and services, from manufacturing robots to driverless cars. A National Algorithms Safety Board, modeled on the National Transportation Safety Board, could make automation as safe as aviation. Fundamental technology solutions are emerging in the form of better audit trails to create a flight data recorder for every robot, more modular designs to support localization of failures during testing and after operational failures, and visualization techniques to better understand how algorithms are actually working.
(2) SHARE BENEFITS: promote policies that equitably share the benefits, by way of higher minimum wages, better medical care, and ambitious efforts to create appropriate new opportunities for displaced workers. Increased access to education and job training with local businesses will help some workers. Other workers may need skills training and encouragement to explore new possibilities like Ebay, Etsy, AirBnB, or Uber. New jobs can help those ready to learn something new: healthcare, equipment maintenance, delivery for Amazon and Fedex, and the leisure, restaurant, and entertainment industries.
(3) LIMIT HARM: anticipate and prevent malicious uses. Research in the Fairness, Accountability and Transparency in Machine Learning community shows promise in preventing biased algorithms and datasets from unfairly withholding parole, jobs, or loans. Bigger challenges remain and even more jobs will emerge in preventing social media manipulation that spreads fake news, promotes hatred, and influences elections.
There is another higher level challenge: we have to remember that human creative capabilities are in a different class from the modest but useful tools of artificial intelligence, such as robots, deep learning, neural nets, and image recognition. Humans make creative contributions by working at the frontiers of knowledge, combining technological building blocks into ever more astonishing innovations. Frontier thinking, which creates valuable new products and services, is a never ending process that improves technology to bring better lives to people.
So let’s stop worrying about job-killing, robot-induced unemployment and get to work on making human-centered automation a success story by ensuring safety, sharing the benefits, and limiting harm.
Ben Shneiderman is a University of Maryland Emeritus Professor of Computer Science and a member of the National Academy of Engineering.
