“Self-aware” robots can learn complex tasks by observing humans. Is that a good thing? : NPR

Imagine a robot that can do your laundry, make your bed, cook dinner, and stock the dairy aisle at your local grocery store. Humans have long been able to teach robots how to perform individual tasks, but despite billions of dollars being invested in robotics, teaching robots to do these more advanced jobs has been an elusive goal.

A team of Swiss scientists is currently making progress towards creating helpful robots that can act on complex instructions from humans. This development raises questions about whether this kind of technology might one day be able to harm humans as well as help them.

Invention of the personal barista

For years, robotics scientist Suthupraya Gupta has dreamed of what his robots could do. “Personally, I’d like to have a robot make my coffee,” says Gupta. He and his colleagues at the École Polytechnique (Switzerland’s engineering school) work late in a lab in the Swiss Alps. “Coffee consumption is high,” says Gupta.

“If you can say, ‘A little more sugar, a little more creamer,’ something like that,” he says. “That would be a dream come true.”

A problem that robotics scientists and engineers like Mr. Gupta have been battling for years is the inability of robots to perform tasks beyond those for which they were specifically programmed. Mr. Gupta uses the example of tennis to illustrate this problem. A robot could learn how to hit a backhand shot, he explains. They can perfectly backhand that ball over and over again. But if the conditions change, for example if the opponent moves or the light changes, everything falls apart. Humans have no problem adapting to such changes. However, teaching robots how to adapt is much more difficult.

“Transferring this behavior from a human to a robot is very difficult,” says Gupta.

I had hoped so until now. Mr. Gupta and his colleagues published a paper in an academic journal science robotics Demonstrate a new way to teach robots using machine learning, a type of artificial intelligence. This approach relies on kinematic intelligence, the robot’s built-in awareness of how its own body can move safely through space.

In a video demonstrating the company’s technology, a one-armed robot attached to a base watches as a human instructor throws a ball into a small container. The robot then picks up the ball, mimics the instructor’s movements, adjusts its position, and adapts to its non-human body. Robots will be able to transfer these skills and knowledge to other robots.

“This could be a tipping point,” said Robert Platt, an engineering and robotics researcher at Northeastern University. Professor Platt, who called the research a “breakthrough,” said that while there is no widespread agreement in the field of robotics about the path forward for developing effective robots through machine learning, most agree that the problem researchers are addressing is a critical one. “More people may do this in the future,” he says.

Pratt was hesitant to predict a specific timetable for robots becoming widespread household accessories. “We are in a period of very rapid change,” he points out. “One of the reasons I’m hesitant to make predictions is look at what’s happened with large-scale language models,” he said, referring to widely adopted generative AI chatbots like ChatGPT and Claude. “We were so far apart and all of a sudden we were so far apart.”

There is a fine line between self-awareness and consciousness

If a robot could self-correct and teach others, would it become self-aware?

“This robot appears to be capable of very impressive learning,” said Susan Schneider, an artificial intelligence researcher at Florida Atlantic University. “But that doesn’t mean that something has full-fledged consciousness or inner awareness in the sense that biological beings have it.”

Schneider points out that the crucial difference between robots and humans is emotion. “Consciousness is the felt quality of experience,” she says. “When I drink my morning espresso shot, when I see a rich sunset, when my head hurts, I feel like something is me from within.”

But this lack of awareness raises new questions about ethics. “This immediately sets off alarm bells in the minds of AI safety researchers,” Schneider says. Future versions of this type of technology could be used as weapons against humans, she says.

Researchers have been careful to include safety protocols to prevent robots from harming people. But even they acknowledge that future developments in this technology require guardrails. “I think there needs to be a regulatory framework in place soon about who can operate robots and how,” says Gupta.

Humans are at a tipping point in robotics, says Susan Schneider. “These are very exciting times, but we don’t know where it’s going yet,” she says.