News

Actions

Magic Leap’s new AI assistant looks alarmingly human

Posted at 9:23 AM, Oct 14, 2018
and last updated 2018-10-14 09:23:09-04

The tech world’s latest virtual assistant looks so realistic, you might mistake her for an actual human.

The tech world’s latest virtual assistant looks so realistic, you might mistake her for an actual human.

Apple has Siri, and Amazon has Alexa. But the lifelikeness of both are dwarfed by Mica: a prototype that Magic Leap, a highly regarded augmented-reality startup, unveiled at its conference Wednesday.

Mica isn’t just a voice assistant. She’s something you can actually see if you wear the company’s augmented-reality glasses, called Magic Leap One. Mica looks and acts like a human — she makes eye contact and offers a warm smile, along with other human-like expressions.

Experts say Mica is a breakthrough in realistic avatars and could have far-reaching implications for society, impacting everything from human connections to education and our health.

For starters, she could be used to turn on or off your smart home devices, including TVs, stereo systems and lights. Magic Leap also demonstrated how Mica can answer detailed questions, such as recalling a person’s favorite song from a concert attended a year ago (the company did not explain how Mica learns such information).

In fact, Mica seems so real that she startled Magic Leap’s test subjects.

“When people came close to Mica they’d instinctively back up, as if they were invading her personal space,” John Monos, Magic Leap’s vice president of human-centered AI, said at the conference. “These incredibly visceral reactions to Mica completely realigned our priorities. Our goal is nothing short of the most realistic human experience in spatial computing.”

According to Monos, when Mica smiled, people smiled back. And when Mica yawned — a contagious social behavior — the subjects would yawn too.

Magic Leap hopes Mica is just the beginning. The company is encouraging computer developers to build similar assistants on its platform.

“We will create a new universe,” Monos said. “Together we will populate that world. We can’t wait to some day meet your Mica.”

This is all possible because of Magic Leap’s breakthroughs in augmented reality, according to Amitabh Varshney, a University of Maryland professor who researches augmented and virtual reality. He pointed to the way light reflects off Mica’s skin, a subtle detail that’s critical to making Mica seem human.

Varshney says the visual realism is closer than ever to leaping over the uncanny valley, a phenomenon in which humans feel discomfort and revulsion looking at something that appears almost human but isn’t.

Magic Leap hasn’t said when it expects people to experience Mica, or an equivalent assistant.

But according to Aditya Sankar, director of research and education at the University of Washington Reality Lab, in a couple decades it could be common for people to wake up in the morning, put on their augmented reality glasses and see their very own Mica.

“If you told someone 15 or 20 years ago [they would have] something in your pocket that finds any information you seek at your fingertips, people would probably not have believed you,” Sankar said of smartphones.

Powerful implications for education — and ethics

The new technology could also potentially impact how we learn.

As Varshney put it: “Wouldn’t it be nice if you walked into the National Gallery and had Van Gogh explain his art?”

Companies might make avatars of historical figures such as Van Gogh, Einstein or Martin Luther King, Jr. Students could sit across a table from a famous figure and get personally tutored on relativity from the person who discovered it. Plus, realistic AI assistants could be able to earn our trust, which could be useful for fields like telemedicine.

This breakthrough raises ethical questions, too.

Avatars introduce the possibility of even more powerful fake news. What if a respected historical figure were used to incite violence? Regulations may be needed to govern the likeness of a person in a virtual world, Varshney said.

But realistic avatars could be used to encourage tolerance. Through augmented reality, a person biased against a minority group could receive a favor from an avatar who looks like a member of that group. In the real world, they would be more likely to reciprocate in a positive fashion toward a member of the minority group, Varshney said.

But critics caution about the hazards of increasingly human-like technologies. We’re hardwired to believe that machines with an avatar or a human sounding voice have empathy, and care for us as a human would.

“Why do we want to attribute personality, empathy, being to an avatar that has none?” said Sherry Turkle, founding director of the MIT Initiative on Technology and Self. “We talk to Siri, Echo, Alexa. We are starting to ask for their advice not just on where to order a pizza but about how we feel about our boyfriends, children, our parents.”

Magic Leap did not respond to a request for comment on that point. Turkle cautions that we’re at an inflection point, similar to when the iPhone was introduced. It was a remarkable innovation and incredibly popular with users — but smartphones can also be addictive and distracting in social situation.

“Tim Cook apologizes and says, ‘Sorry! Here are tools to walk it back,'” Turkle said. “What do we forget when we talk to machines? We forget what it is to be human — what is special about talking to a human. That is the danger.”