Google: Useful artificial intelligence finally here

Spring may finally have arrived for artificial intelligence, Google executives said Friday.

But summer remains a ways off.

Speaking at the Google I/O developers conference in Mountain View, executives said that artificial intelligence and machine learning have advanced to the point where they are proving genuinely useful, through such technologies as speech recognition and language translation. But there remains great room for improvement.

“We’ve seen extraordinary results in fields that hadn’t really moved the needle for many years,” said John Giannandrea, vice president of engineering for Google. “I think we’re in an AI spring right now.”

This year, the company’s AlphaGo became the first computer program ever to beat a professional player at Go, the ancient strategy game considered more complicated than chess. AlphaGo beat Go master Lee Sedol four out of five games. Google CEO Sundar Pichai called it a game changer, pun presumably intended.

“This is another important step toward creating artificial intelligence that can help us in everything from accomplishing our daily tasks and travels to eventually tackling even bigger challenges like climate change and cancer diagnosis,” Pichai wrote in an April letter to shareholders of Google’s parent company, Alphabet.

The company believes that there is room for developers to use machine learning to tackle ever more complex challenges. Jeff Dean, a senior fellow in Google’s research group, said that if image recognition technology improves to detect fine details in photos, it may help solve health problems such as detecting diabetic retinopathy, which can cause blindness.

Aparna Chennapragada, a director of product management at Google, said she believes there is also opportunity for using machines to help the elderly. Instead of an 80-year-old woman with a bad back having to pick up something off the floor, a robot could do it for her, she said.

“You want to look at problems that are easy for machines, hard for humans,” she said.

But Google executives said that artificial intelligence still has a ways to go.

Giannandrea said he wouldn’t declare that AI has reached “summer” until machines are able to read and internalize text, understanding the context enough to paraphrase it. He also noted that while Google’s AlphaGo was great at the game Go, the program isn’t able to translate what it learned from Go into, say, playing a game of chess.

There have also been some hiccups in machine learning. Last year, Google Photos and Yahoo’s photo-sharing service Flickr, for example, miscategorized photos of black people as gorillas or apes.

It’s also unclear how machine learning will be used commercially. Google on Wednesday unveiled a voice-activated, Web-connected device called Google Home, a competitor to Amazon Echo. Giannandrea said that in general, Google focuses on building products that are useful then finds a way to make money from them if they become popular.

Laurent Lec, a software director from SoftBank Robotics Europe, on Friday encouraged developers at I/O to create programs for his company’s humanoid robot, Pepper. The robot already is used to greet shoppers at SoftBank stores in Japan — at one point, Lec said, Pepper robots ran a shop for a week. Pepper is being shipped to the U.S. this year.

Lec said robots of the future “would basically do things that humans don’t want to do or cannot do.” Earlier in the conference, he showed how Pepper, a white robot with a human-like face, could read his hand signals and how, through a VR headset, he was able to see through Pepper’s eyes into the audience.

“It’s a bit disturbing,” he said, because it made him feel like he was somewhere else. “My senses are connected to a different place.”