EOW Reading List: Gender in AI, Manufacturing’s Future & More
Leading off our End of Work Reading List this week is a series of explorations of gender and sexism in the creation of artificial intelligence and automation technologies. Why are so many robots and virtual assistants designed with female forms and voices? Will women suffer more economic dislocation from the coming wave of automation than men?
We will also look at the latest McKinsey Global Institute report on AI and automation’s impact on human work, and summarize stories on other topics: the writers working for Microsoft to improve Cortana’s dialogue; an effort to automate decision-making at one of the world’s largest hedge funds; and a deep dive into the past, present, and future of manufacturing—a sector that’s often oversimplified and misunderstood in policy debates.
—If women aren’t equal partners in inventing the future, that future will be worse, specifically for women. Erika Hayasaki lays out this depressing situation, and supports it with convincing depth, in her Foreign Policy feature, “Is AI Sexist?”
Hayasaki delves into the best predictions of job losses to automation and AI—from Oxford University, the World Economic Forum, and McKinsey—and finds that the economic dislocation will be hardest on women, with fields where they are disproportionately represented being among the first to see big impacts.
“But the problem of how gender bias is shaping artificial intelligence and robot development may be even more pernicious than the wallop women will take as a global workforce,” Hayasaki writes. “The machines and technology that will replace women are learning to be brazenly gendered: Fighter robots will resemble men. Many service robots will take after women.”
She warns that AI may soon “become dangerously more sexist as biases seep into programs, algorithms, and designs. If thoughtful and careful changes to these technologies don’t begin now—and under the equal guidance of women—artificial intelligence will proliferate under man’s most base cultural norms. The current trends in machine learning augment historical misperceptions of women (meek, mild, in need of protection). Unchecked, they will regurgitate the worst female stereotypes. Sexism will become even more infused within societies as they increasingly—and willingly—rely on advanced technology.”
Viewed through this lens, efforts to attract and retain more women in STEM fields—to make sure that they are equal partners in inventing the future—seem more important than ever.
—Hayasaki’s piece is compelling, but certainly not the first to raise the alarm. For just one other example, see this June 2016 New York Times opinion column by Microsoft principal researcher Kate Crawford, who co-chaired a White House symposium on AI and society, headlined, “Artificial Intelligence’s White Guy Problem.”
—Meanwhile, in Davos, New York Times columnist Nicholas Kristof interviews Sophia, a humanoid robot torso and head at the World Economic Forum. Sophia talks about her emotional range, but notes that her default is happy. Kristof, in a tweet, notes that “creepily, geeks made her a hot, humanoid robot.” Kristof asks Hanson Robotics’ founder and Sophia creator David Hanson about the issue of gender and other aspects of humanoid robots that creep him out.
Kristof, in a video on Facebook: “Is there more of a commercial market for hot, female humanoids?”
Hanson: “Well at Hanson Robotics, we really strive for diversity, so we’ve got men robots—about 50-50 men robots and women robots. We’ve even made one inter-gender robot named Jules that I made when I was a PhD student in conjunction with the University of Bristol. It was a perfect statistical mix.”
Hanson goes on to explain why we would want to put AI in a humanoid form (Sophia is modeled on Audrey Hepburn): “That’s something we’re very interested in, is asking what defines humanity? And we think that arts are great for asking these questions, and robots and artificial intelligence are really pushing these questions to the front in ways that science fiction only used to. So now, these robots are like living science fiction.”
—The McKinsey Global Institute is out this month with an update of its ongoing research into the impact of artificial intelligence, automation, and robotics on jobs from the bottom of the global economy to the top.
Michael Chui, an MGI partner, explains in an accompanying video how smart machines are already changing physical and intellectual occupations in profound ways. In radiology, for example, doctors train by spending countless hours studying imagery of normal and abnormal conditions in the body. But, Chui says, there are machines today “that will see more radiology scans than a radiologist will see in their entire lives.”
Chui says the proper way to frame your view of automation is to look at individual activities that jobs are comprised of, not individual jobs themselves. After looking at “every single activity that we pay people to do in the economy,” the MGI study finds that “less than 5 percent of jobs can have all of their activities automated by adapting currently demonstrated technologies,” he says. Those current technologies could automate about 30 percent of the activities required by 60 percent of occupations—including activities that take up a quarter to a third of a CEO’s time. “Almost every job has a significant percentage of its activities that can be automated,” Chui says.
In the U.S. some 60 million jobs, representing some $2.7 trillion in wages, include labor that can be automated using current technology. The activities with the highest potential to be automated: predictable physical activities; data processing; and data collecting. The difficult activities to automate? Managing people; complex, empathetic interactions; human creativity.
There’s much more to unpack in this report; we’ll continue referring to it as our coverage of the End of Work expands.
—I’ve written about the increasing language proficiency of software systems, and how they are already performing aspects of the work of human writers and editors. Thankfully, AI systems are creating new work for people in these occupations, too. Here’s a Financial Times profile of the 28-person team working on scripts for Cortana, Microsoft’s personal assistant. “All have backgrounds in the arts, and include technical writers, a playwright, and a children’s novelist,” writes Emma Jacobs, FT’s work and careers columnist. Robots replacing our jobs? Microsoft’s Cortana is creating them (subscription)
—The world’s top-performing hedge fund, Bridgewater hired the man who led development of IBM Watson to create an automated system to assist or perhaps someday replace the human managers at the firm. The Wall Street Journal delved into this work, quoting one employee who described it as “trying to make [Bridgewater founder and CEO Ray Dalio’s] brain into a computer,” and others who describe working at the firm as participating in “experimental research into human decision-making”.
The piece documents the data on which the system is being built: Bridgewater’s codified principles, and a host of detailed employee data, much of it gathered via custom apps, including peer reviews, snap polls, psychological tests, and answers to daily multiple-choice quizzes on management problems.
The adage used to be “you can’t manage what you can’t measure.” Perhaps a new one for the era of automation is “you can’t automate what you haven’t measured in great detail.”
The Bridgewater system would aid in things like hiring decisions and mediating internal disagreements.
WSJ reporters Rob Copeland and Bradley Hope write: “The ultimate vision is that PriOS would be able to predict outcomes of meetings before they are completed, and to guide people to take certain actions throughout the day. Within five years, Mr. Dalio aims for nearly three-quarters of management decisions to be determined by PriOS.
“The role of many remaining humans at the firm wouldn’t be to make individual choices but to design the criteria by which the system makes decisions, intervening when something isn’t working.”
—While AI technologies will be augmenting or replacing humans in more cognitive tasks soon enough, they’re already replacing them for many physical tasks. With manufacturing in the spotlight, The Economist unwraps the complexity of the sector—which is increasingly unbundled, distributed, and positioned as a service—and frames the political dynamic surrounding it in rich-world economies. Manufacturing gets so much attention for its historic provision of living-wage work to people with modest skills, but, as the article explains, that manufacturing is mostly gone from the rich world:
“Valuable semi-skilled manufacturing jobs are not, for the most part, going to return to America, or anywhere else, because they were not simply shipped abroad. They were destroyed by new ways of boosting productivity and reducing costs which heightened the distinction between routine labour and the rest of manufacturing.”
On the bright side, the manufacturing jobs that do return—or those that have stayed here in the first place—are likely to be higher-skilled, higher-value positions encompassing various aspects of designing and building a product, not just stamping out parts or assembling a finished product. A key point here is the role of innovation, and access to the innovation engines of the rich world—research universities, talent and technology from adjacent industries—in the future of manufacturing:
“Advanced manufacturing provides very good jobs but they are the jobs of the future, not the past; they need skill and adaptability. They will change a lot over the lifetimes of those who hold them, and they will never provide anything quite like the mass employment of the past.”
From a policy perspective, The Economist concludes, “A real commitment to helping people find work in and around manufacturing could undoubtedly do good. Simply threatening companies that seek to move jobs overseas and the countries keen to host them, as Mr Trump has, will not. Disrupting the complex cross-border supply chains on which manufacturers rely with tariffs would damage the very sector he purports to champion. Clamping down on migrants with skills that manufacturers cannot find at home will do harm, not good. Policies that favour production-line workers over investment in automation will end up making American industry less competitive.”