Why Robots Aren't Going To Ride To Our Rescue

Turns out humans are pretty useful - and frequently more competent than robots -  especially when it comes to tricky stuff like treating patients, as opposed to just diagnosing them. JL

Matt Simon reports in Wired:

One in five US households has already lost work because of the vulnerabilities of the human worker. When we get sick, work that depends on people grinds to a stop. Why haven’t the machines saved us yet? Our economy craters without human workers because the machines are far away from matching our intelligence and dexterity. (And) we may never be able to automate the industry that needs it the most: medicine. A hospital isn’t an automotive factory floor; patients stricken with this disease are severely ill and scared as hell. Would you task a machine with deciding who lives and who dies?

MODERN CAPITALISM HAS never seen anything quite like the novel coronavirus SARS-CoV-2. In a matter of months, the deadly contagious bug has spread around the world, hobbling any economy in its path. In the United States, where consumer spending accounts for more than two-thirds of economic activity, commerce has come to a standstill as people stay home to slow the virus’ spread. Hotels and restaurants and airlines have taken massive hits; Delta has cut its flight capacity by 70 percent. One in five US households has already lost work. And that’s all because of the vulnerabilities of the human worker. When we get sick—or we have to shelter in place to avoid getting sick—the work that depends on people grinds to a stop.

Why haven’t the machines saved us yet?

This economic catastrophe is blowing up the myth of the worker robot and AI takeover. We’ve been led to believe that a new wave of automation is here, made possible by smarter AI and more sophisticated robots. San Francisco has even considered a tax on robots—replace a human with a machine, and pay a price. The problem will get so bad, argue folks like former presidential candidate Andrew Yang, we’ll need a universal basic income to support our displaced human workers. (UBI seems to have actually arrived, in a sense, with the Trump administration’s proposed payout to American households to weather the crisis: A $1,000 check for most, with an extra $500 for every child.)

Yet our economy still craters without human workers, because the machines are far, far away from matching our intelligence and dexterity. You’re more likely to have a machine automate part of your job, not destroy your job entirely. Moving from typewriters to word processors made workers more efficient. Increasingly sophisticated and sensitive robotic arms can now work side-by-side on assembly lines with people without flinging our puny bodies across the room, doing the heavy lifting and leaving the fine manipulation of parts to us. The machines have their strengths—literally in this case—and the humans have theirs.

“Robots can very successfully augment human activities,” says Julie Carpenter, a roboticist and research fellow at the Ethics and Emerging Sciences Group at Cal Poly San Luis Obispo. “They can do the labor we don't want to do or can't do, and are especially successful at carrying out tasks that we consider repetitive, boring, or dangerous,” like lifting car doors on an assembly line, for example.

But they’re not very smart, especially when it comes to problem-solving. Think about how you would pick up a piece of paper that’s lying flat on a table. You can’t grip it like you would an apple—you have to either pinch it to get it to lift off the surface, or drag it to hang over the edge of the table. As a kid, you learn to do that through trial and error, whereas you’d have to program a robot with explicit instructions to do the same.

During the pandemic, this contrast between humans and machines has become particularly fascinating in Amazon’s warehouses. Last week, Amazon officials announced that in response to the coronavirus they were hiring 100,000 additional humans to work in fulfillment centers and as delivery drivers, showing that not even this mighty tech company can do without people. But it, too, is automating parts of jobs. In one warehouse near the Denver airport, the company has deployed squat little robots that ferry packages between human workers—doing the heavy lifting while leaving the fine manipulation of objects to people.

Amazon’s automation technology will only get better from here. But the demand for products will keep climbing as well, as we’re seeing with this hiring bonanza. “Their need for human labor may fall through time, but for now the growth in demand for their products outstrips any gains from automation,” says Dean Baker, senior economist at the Center for Economic and Policy Research and visiting professor at the University of Utah.

And if longer-standing industries are any indication, the machines will need human coworkers for some time. “Even the heavily automated industries still rely on humans for essential tasks,” says Baker. North America’s auto plants just shut down, not because robots can catch coronavirus but because their human operators can. Even when San Francisco Bay Area counties instituted a strict shelter-in-place order, Tesla claimed its Fremont factory was an essential business and must remain open. To which the county sheriff said, yeah, nice try, and ordered the plant shut.

Actual essential businesses, like grocery stores, remain open. But while Amazon is hiring, other industries are cratering in this crisis, mostly ones that rely on gathering people together in one space. Hospitality workers have been particularly hard hit, as bars and restaurants and hotels have closed; 4.6 million people in the travel industry could lose their jobs. Hospitality is the operative word here. There’s a reason you don’t see too many robot bartenders—well, two reasons, actually: Robots can’t match our manipulation skills, and no one goes to a bar to banter with a machine. Even though Silicon Valley has been obsessing as of late over robots that make pizza and coffee and burgers—basically, restaurants where you’re not bothered with pesky human interaction—pretty much all of them had cratered even before the coronavirus crisis.

“We know that robots are great at certain things right now, like repetitive work,” says Carpenter. “And they can do that forever. What's not so great is anything that has to do with a human-centered context, a cultural context.”For example, we may never be able to automate the industry that now needs it the most: medicine. Doctors and nurses and other health workers around the world are working themselves to exhaustion, and many are falling ill. A hospital isn’t like an automotive factory floor; bedside manner matters. Patients—especially those stricken with this new disease—are severely ill and scared as hell. “Whether it's physical or emotional, people need to feel like their pain is being heard, that their implicit suffering is made explicit and have that reflected back to them,” Carpenter says. Good luck teaching a machine to empathize with a human on the brink of death. And indeed, it’s this empathy gap that makes many roboticists think that for this reason alone, we shouldn’t automate other particularly sensitive jobs, like police work and education.

It’s not even clear if we could make robots and AIs sophisticated enough to handle the job. In Italy, overwhelmed hospital staffers have been making devastating triage decisions, prioritizing those that can be saved. But would you ever task a machine with deciding who lives and who dies? Hell, we’re having a hard time even figuring out how the algorithms that run autonomous cars should make safety decisions, and that’s divorced from the much larger scale of coronavirus triage.

“I hate to say this, because it's so overdone, but it's like the trolley problem, times billions,” says Carpenter. The trolley problem is a classic thought experiment: If you’re driving a trolley and you see five people on the track ahead, but only one person on a side track, do you pull the lever to switch tracks, actively making a decision that will kill one person? Or do you do nothing, and passively let the five others die? These are not decisions we want machines to be making when it comes to hospital care. “Robots are great at augmenting human skills,” says Carpenter. “But as the situation now is clearly demonstrating, we absolutely very much need humans to understand larger situations and make decisions.”

Still, robots can go a long way toward assisting humans in a hospital, but they’re particularly tricky to design. Consider an autonomous robot named Tug, which wheels through the corridors delivering drugs to nurses and food to patients. Its creators went to great lengths to make it helpful yet polite. It gently beeps so humans know it’s nearby, and when it waits for elevators, it tells you so: “Waiting for doors to open.” Some hospitals dress it up in costumes to amuse kids. It’s a tool, not an employee in and of itself. By helping deliver supplies, it frees up nurses to do what they do best as humans: interact with patients.

Overestimating robots and AI underestimates the very people who can save us from this pandemic: Doctors, nurses, and other health workers, who will likely never be replaced by machines outright. They’re just too beautifully human for that.