TOKYO ― Google is rethinking everything with machine learning, an artificial intelligence technology that trains computers to think and learn from data. 


“Machine learning has actually been around for a long time. But it has made huge achievements over the past three years,” Eric Schmidt, executive chairman of Google’s parent company Alphabet, said via video.

“Google is the world leader in the field. Machine learning will make Google even smarter.”

Machine learning is based on deep learning theories that are inspired by how human brains work. The software is trained to handle vast amounts of data and learn from experiences and errors.

Google, the world’s largest search engine, now uses the technology not just for search and ad targeting but also for computer vision, language translation and facial recognition.

Touting itself more as a machine learning company, the company is pouring considerable resources into related R&D activities and people.

Its latest Smart Reply technology creates responses to emails in a user’s Gmail inbox. The option reads the content and suggests short phrases at the bottom of the screen which you can use to reply.

Google Photos, launched in May, has built-in facial recognition that can automatically organize photos based on people’s faces. But the “face-grouping” feature is currently available in the U.S. only.

Google Translate, among other things, attracted more attention from the Asian journalists during the event as they showed keen interest how well the software works with diverse languages other than English.

Users can type or take a photo of text with the app and then the technology looks for patterns across hundreds of millions of human-translated documents and makes an intelligent guess about which translation makes the most sense.

“The software may respond better to more common languages for now,” said Greg Corrado, Google’s senior research scientist. “But it is getting better and better. Computers are slow learners. They have to look at many examples to get insight.”

Based on a demonstration session, the app’s recognition was accurate enough to help users easily engage in casual conversations between Korean and English. The machine learning technology’s translations are expected to improve as more speakers use the software.

Data collection is the key to machine learning. And Google also understands privacy concerns its users feel in an ever-evolving Internet era.

“We work very hard to keep people’s data private and to give you control over it,” Schmidt said, refuting complaints about lax privacy protection measures by IT companies.

“To my knowledge, the safest place to put your data is inside Google. It’s far safer than, for example, giving it to the government.”

Google is also seeking to expand ties with other developers and companies in its latest push for machine learning.

The company this week announced it would make its latest machine-learning program TensorFlow available for free as an open-source software. By releasing the software, the company aims to make it a standard toolset used by researchers and create a Google-led ecosystem across industries.

The Alphabet chairman also hinted at an enhanced partnership with Korean firms on the field during his recent visit to Seoul.

By Lee Ji-yoon (jylee@heraldcorp.com)