Watch (3:23)

An Antidote to Automation Anxiety: Intelligence Amplification

Ken Goldberg is not one for robot panic. At What’s Now: San Francisco, the William S. Floyd Distinguished Chair of Engineering at UC Berkeley delivered a fascinating and thoroughly convincing presentation outlining why we should be excited about, and not afraid of, our robot-filled future.

For starters, Ken thinks the threat of robots and artificial intelligence (AI) stealing large swaths of jobs is overblown. “I don’t think we’re going to see fully autonomous cars on the road in the foreseeable future,” Ken said, stressing that autonomous driving is a very difficult problem that requires tremendous amounts of data. He believes the hype surrounding AI is driven at least partly by investors and companies, like Uber, that are setting overly ambitious deadlines for the release of technologies like fully autonomous vehicles. Ken pointed out challenges that even the most advanced robots have with seemingly simple tasks, like grasping objects and classifying images, and how tiny alterations that wouldn’t faze a human can render a robot incapable of performing its assigned task.

In addition, we shouldn’t worry about robots “replacing” us, Ken insists—rather, we need to focus on how we can work together with machines to improve efficiency and performance. “Computers should have a seat at the table,” Ken said. Technology has dramatically transformed education in the past, when advances in agriculture gave more Americans time to pursue high school degrees. In 1910, only 10 percent of Americans graduated from high school. By 1950, 80 percent of Americans completed high school.

Given this precedent, Ken encouraged the audience to view AI as an opportunity to revamp the way we teach, and incorporate more creativity and innovation into the educational system. The things that computers are best at—calculations, precision, and objectivity— are distinct from the qualities that (at least for the time being) belong primarily to humans—understanding, purpose, and passion. Ken believes these complementary abilities portend a future driven by not AI, but IA—intelligence amplification.

According to Ken, the source of automation anxiety is similar to the source of anti-immigrant sentiments: people want jobs, and if they don’t have them, they historically have blamed immigrants. These days, there’s a new scapegoat in town: robots. “Robots are immigrants,” pointed out Oliver Morton at the Economist. According to Ken, these “immigrants from the future” can be viewed either as a threat or an opportunity, depending on whether you’re on Team Elon Musk or Team Mark Zuckerberg.

Ken, who’s firmly on Team Zuckerberg, used the example of “random forests” to illustrate his point. Multiple decision trees perform better than a single one, Ken noted, but the key is the trees have to be sufficiently different. This analogy also applies when it comes to groups of humans—heterogeneous groups have a higher collective IQ than homogeneous groups. “Diversity in hiring is not about correcting a historic wrong or being politically correct, it’s about getting better performance.” Ken said. Robots working together with humans, amplifying our intelligence, is yet more one type of diversity—one that, if we don’t let fear or prejudice stand in the way, has the potential to deliver as of yet unimagined benefits.

As another example of a field in which robots “aren’t quite there yet,” Ken shared recent research in the field of robotic surgery. Intuitive Surgical has 3,000 robots in operating rooms, but each one is controlled by a surgeon, Ken said. While these robots are not yet autonomous, and are nowhere close to stealing jobs from surgeons, Ken and other roboticists are currently conducting experiments to try to make robot more autonomous in order to better assist surgeons. To do this, they take data from human surgeons and infer models from these examples, attempting to teach robots underlying techniques that are consistent with all of the demonstrations. The robots that Ken and others are working with still have a 20 percent failure rate, Ken pointed out. He believes that roboticists should be more honest about their “blooper reels,” rather than trying to convince the public of the imminence of a future that Ken believes is still very far away.

Ultimately, Ken argued, we need to choose whether to view a future of increasing automation and AI with trepidation and anxiety, or with hope. Ken presented a pretty convincing case for the latter attitude. Realizing his optimistic vision for Multiplicity—humans and robots working together—will require rethinking the way we teach and work, and a renewed commitment to diversity of all kinds.