SAN FRANCISCO — Willie McTuggie looks like a photocopier on wheels. But he — it, actually — has the engineered brain of a reasonably smart human, and acts like one when he rolls up to a nurse’s station, opens a drawer, retrieves a dose of pills and glides off to make a delivery.
Packed with more than 30 motion-detecting and other sensors, Willie and his automated buddies at the UCSF Medical Center can open doors, avoid collisions with doctors on rounds and perceive when to wait for a free elevator.
There are 25 mobile bots from robotics company Aethon on staff, named and decorated by mortal colleagues. Willie is wrapped in the San Francisco Giants’ team colours of orange and black, and Maybelle is designed to look like one of the city’s cable cars. They perform duties once handled by nurses, orderlies, cafeteria staff and maintenance crews, though so far no people have lost jobs to the bot corps.
“It does displace certain roles, but we can put that head count into other service roles,” says Pamela Hudson, executive director of clinical systems at the University of California, San Francisco, hospital. It is, she says, a win-win.
Not everyone is enthusiastic as contraptions and software coded with artificial intelligence invade the workplace. The human-brain mimics are becoming so clever that, according to a study by the Oxford Martin Program on Technology, 47 per cent of all U.S. jobs are at risk over the next two decades of being given over to computers.
They’re already writing sports stories, milking cows and reviewing X-ray results. Three-foot-tall cybernetic bellhops invented by Savioke, a robotics company, deliver room-service orders at Aloft hotels near Apple’s headquarters wearing painted-on black bow ties. The startup Momentum Machines is building a fast-food burger-flipping apparatus. At the University of Maryland Institute for Advanced Computer Studies, a Baxter robot from Rethink Robotics is mastering the art of making a salad.
The artificial intelligence revolution is writing a new chapter in the age-old debate over whether machines are putting people out of work or opening up new opportunities for them. “The idea of technology destroying jobs has been going on for two centuries,” says Richard Cooper, an economist at Harvard University who has studied the impact of technological advancements on employment. “Certain jobs get destroyed but other jobs get created.”
The catch in the 21st century is that the technological leaps are so big and happening so quickly, and at a time when service industry jobs are responsible for more than 40 per cent of employment growth in the U.S., where income inequality is widening.
“The bar to get entry in to the labour force is rising faster than people expected and the ability to stay there is falling,” says Sebastian Thrun, former head of the Google research laboratory Google X and a developer of the company’s driverless-car technology. “The competition from machines is getting stronger and stronger.”
That’s because they’re getting smarter and smarter. Super-fast computer-processing strengths and the information-scavenging abilities of the web make it possible for machines to quickly process huge amounts of information, learn from it and share — like when a self-driving car is in a fender-bender after going too quickly around a turn and transmits a warning to others so they don’t make the same mistake.
In so-called deep learning AI systems, tens of thousands to millions of digital neurons are stitched together and layered to create a Frankenstein version of our own neocortex. These can learn about data merely by being exposed to it, and are already widely used in cutting-edge digital imaging. At Facebook, researchers are designing software that can read simple texts and answer questions about it. At Google, engineers have built systems that allow a computer to absorb the rules of an arcade game, learn to play it and win.
Last month, Google received a patent for instilling a robot with a personality tailored to mesh with the human with whom it’s interacting — or, as the patent put it, display “states or moods representing transitory conditions of happiness, fear, surprise, perplexion (e.g., the Woody Allen robot), thoughtfulness, derision (e.g., the Rodney Dangerfield robot), and so forth.”
That future isn’t quite here yet. Androids on the payroll have varying levels of smarts and sophistication. Some, like Willie McTuggie, are loaded with navigation cunning that can follow programmed maps of a facility to get around. Others attain a refined level of dexterity and understanding of space, which is enough to replace workers on a factory floor.
Bots have been helping assemble automobiles in Detroit for decades, and other manufacturers are enlisting them to perform increasing complicated duties. In Seattle, Boeing’s planning to have KUKA AGautomatons fasten the fuselage panels of its 777 and 777X planes. The bots will handle the drilling and filing of more than 60,000 panels, which according to the aircraft maker will boost worker safety and product quality. Workers on the fuselage will transition to new roles, according to Boeing.
Meanwhile, at the University of Maryland, the Baxter robot — named Julia, after chef Julia Child — watches cooking videos on YouTube and learns, step by step, what to do. The magic is in the bot’s brain, which is loaded with advanced image-classification software and a reasoning system that translate what it “sees” through cameras positioned on pincher claws at the ends of its two big red arms. Julia observes and then pours lettuce and baby tomatoes in to a bowl, adds dressing, and then imitates how a chef’s hand grasps a spoon to mix the concoction.
Julia is years away from taking over as a line chef, but lawyers are already feeling the brunt of deep-learning advances: Software is capable of scanning documents and emails to figure out what’s admissible in trials. “What used to take a hundred attorneys can now be done with one,” says Andy Wilson, chief executive of Logikcull, which used to be a paralegal-for-hire company and now sells legal automation technology.
These days AI teams are working on systems to put some of their own out of work, as Google researchers experiment with systems that can automatically check the quality of a program’s code. “We don’t live in a world where any job last forever,” Thrun says. With technology advancing so swiftly, “people have to keep running.”
Bruce's Blog 
