Computers That Listen And Learn Better Than You Do

From suggesting restaurants to diagnosing patients, computers seem to be getting smarter each passing day. The New York Times looks at some of the powerful technology coming to light, as computers begin listening and reacting to our problems.

The New York Times piece begins with a mother and son entering a room, where a medical assistant helps them diagnose the young boy’s diarrhea issues, an embarrassing process for a person of any age to endure. The personal assist looks down at the boy, asking him if his tummy has been hurting. The boy responds in the affirmative. A few questions later, and the assistant schedules a doctor’s visit for later in the week, telling the mother and her child that she’s not particularly concerned about the boy’s condition.

This sort of scene plays out at doctors’ offices around the country on a daily basis, but this doctor’s assistant isn’t real. She’s a computer. A disembodied face on a screen, powerful enough to recognize speech, tell the difference between a mother and child, and access a database of information quickly enough to make an on-the-spot diagnosis.

She is the future.

“Our young children and grandchildren will think it is completely natural to talk to machines that look at them and understand them,” said Eric Horvitz, a computer scientist at Microsoft’s research laboratory who led the medical avatar project, one of several intended to show how people and computers may communicate before long.

The key to computers that communicate lies in making computers not only listen to what we are saying, but to understand it as well.

Voice recognition software has been around for decades, but it’s only recently that it’s become significantly advanced enough to catch the subtle nuances of the human voice.

According to the article, the number of doctors using speech recognition software to transcribe notes has tripled over the past three years to 150,000, which is a huge number, especially considering the number of obscure terms doctors use on a regular basis.

Once speech recognition is mastered, the sky is really the limit. Translation programs are available that can translate two languages back and forth on the fly, already in use by the military in Iraq.

At Microsoft, Eric Horvitz’s office is guarded by a virtual receptionist. Using a combination of speech recognition and the analysis of Horvitz’s phone calls, appointment calendar, and other data, the receptionist can tell a visitor how long they’ll have to wait to see him d what activity he’s engaged in at the moment. If he’s on the phone with someone who he’s historically had short phone calls with , for instance, the receptionist might tell the visitor to wait five or six minutes.

Advanced features even allow the receptionist to make small talk. If she recognizes the visitor, she might bring up past appointments, or mention events both the visitor and Horvitz attended.

The technology gets more complicated and more refined with each passing day, but unlike humans, practice doesn’t always make perfect.

Take the virtual personal assistant Siri, which returns restaurant suggestions after analyzing voice requests from iPhone users.

Nelson Walters, an MTV television producer in New York, is a Siri fan. It saves him time and impresses his girlfriend. “I will no longer get lost in searching Yelp for restaurant recommendations,” he said. But occasionally, Mr. Walters said, Siri stumbles. Recently, he asked Siri for the location of a sushi restaurant he knew. Siri replied with directions to an Asian escort service. “I swear that’s not what I was looking for,” he said.

As funny as that example might be, it shows that there’s a long way to go before we let learning, listening computers handle the more important aspects of our lives, like medical care.

“It’s not human intelligence, but it’s getting to be very good machine intelligence,” said Andries van Dam, a professor of computer science at Brown University. “There are going to be all sorts of errors and problems, and you need human checks and balances, but having artificial intelligence is way better than not having it.”

Computers Learn to Listen, and Some Talk Back [The New York Times]