Julia Lundy is blind. She uses a wheelchair whenever she ventures more than a few feet. When Lundy began attending classes at the University of Georgia last fall, her mother went along to help her navigate campus life. Naturally, Lundy wanted more independence. She approached Dr. Don Potter, director of the Artificial Intelligence Center at the University, to see if he could help.
Potter and graduate students in the AI Center worked with Lundy to assess her needs. Potter said that at first, they planned to have Lundy’s wheelchair do a lot of work for her, “so Julia could say, ‘I’m in Don’s office; take me to the elevator.'”
But Lundy didn’t want that. She wanted faster and better feedback so she could maintain control. Balancing Lundy’s need for independence with her need for machine assistance is a continuing challenge for Potter and his students.
Medical researchers have experimented with various forms of AI systems for more than 30 years. While Lundy’s situation proves that relationships between people and machines can take time — lots of it — to develop, another situation is driving rapid acceptance of AI and machine caretakers.
The baby boomers are aging.
The universal shortage of nurses in particular and healthcare workers in general is getting worse. Meanwhile, baby boomers are moving like a tidal wave toward assisted-living homes. Soon, patients could face the choice of being cared for by a robotic nurse or no nurse at all — which might be better than it sounds.
When Pearl, a robotic nurse developed by researchers at Carnegie Mellon and the University of Pittsburgh, rolled into service at Pitt’s School of Nursing and later into a retirement community in Oakland, “people loved it,” Sebastian Thrun, associate professor of robotics computing at Stanford University, told TechNewsWorld. Thrun was part of Pearl’s development team at Carnegie Mellon.
“Robots and people are quite complementary in many ways,” he said. This might be especially true for elderly people. “The initial focus [for Pearl] was on helping the homebound elderly and on creating a vehicle for remote healthcare uses.”
Predictably, he said, the researchers moved the project into nursing facilities, having Pearl escort patients through the facility and remind them to take their medicines. “The entire place was really into it,” noted Thrun.
Practical, Untiring Help
Certainly, some of Pearl’s popularity stemmed from sheer novelty. Novelty aside, however, Pearl offered very practical help. In some ways, that help was better than what could be offered by a human counterpart. Consider that an elderly patient using a walker travels at a speed of about five centimeters per second. While human escorts get very impatient, Thrun said, robots adapt to your speed.
Noting that some find the idea of robot conversation distasteful, Thrun pointed out that robot communication offers certain advantages over human interaction. Human nurses caring for dementia patients understandably become frustrated repeating one phrase 10 to 20 times.
A robot, on the other hand, can say “please take your medicine” 200 times without flinching. And a smart robot can say “please confirm that you took your medicine,” prompting the patient to use a touch-screen display to indicate that the medicine indeed went down. The robot also can record and send information to a nurse’s station for human eyes and human follow-up, if necessary.
As more AI functions are built into Pearl and other robots like her, they will offer even greater benefits to the elderly, Thrun said.
Two primary reasons why older people are institutionalized are danger of falling and the inability to manage bathroom needs. Monitoring a person who needs bathroom reminders or who is in danger of falling requires round-the-clock vigilance, which is impractical and expensive for humans. However, 24-hour vigilance is trivial for a robot, noted Thrun. “There’s a whole universe of things that are possible because the robot is available to you 24-7.”
InTouch Health, a development company based in Santa Barbara, California, has placed four robots called “Companions” in nursing facilities around the country. The units lease for about US$3,000 per month, a price that includes tech support.
“Our goal is to leverage the capabilities of a doctor, a nurse or a caregiver,” Marco Pinter, director of video and AI technologies at the company, told TechNewsWorld. “You just can’t hire enough healthcare workers; [The Companion] is helping them be more efficient.”
Today’s Companions employ very limited AI functionality, Pinter said. The Companion’s operator — usually a doctor or nurse — uses a joystick at a computer station to send the Companion to a patient’s room. Once in the room, the Companion’s screen, which displays the driver’s face in real-time, uses full pan-tilt-zoom capabilities to address multiple people in a room or to zoom in for a close look at a wound, Pinter said.
This capability produces an interesting psychological side effect. “When the driver moves the Companion’s head around, people in the room respond as they would in a normal conversation,” he said. “They turn their heads to talk to it.”
Pinter noted that upgrades being tested now include a stethoscope and an oximeter, which would allow Companions to gather physical data in addition to visual and audio input.
“This is coming in layers,” Thrun said of AI’s increasing role in medical care. “The next step is to help manage a person’s life … for a machine to learn a person’s habits on its own.”
It might be difficult for an artificial caregiver to learn a person’s habits about leaving the stove on, or to know that if a patient’s favorite TV show is on at 8:00, that the patient should be taken to the bathroom at 7:45. But this capability seems to be the next milestone to target in medical artificial intelligence. While technology is improving rapidly, Potter believes more improvement is needed before AI medical applications move into mainstream patient care. “We need better sensors and better processing capabilities,” he said.
Science has yet to design a machine that works as well as the human eye in transmitting information to the brain. “It’s getting better every day,” he said. “I can imagine that, by the time I need a wheelchair, the chair might know that it’s 8:00 a.m. and it will take me to the cafeteria for breakfast.”
Initially, the thought of robotic nurses taking care of us as we age might seem bleak. But faced with a serious shortage of healthcare workers, we might have to accept the idea. And, Potter noted, “people can get sick, bored, or not come to work.” All of which makes AI’s less-than-human qualities look pretty good.