Summit: Day 2
"The Future of A.I. Forum"
WEDNESDAY, May 24 overview
PRESS RELEASE
GOOGLE, DEEPMIND, CHINESE RESEARCHERS AND ENTREPRENEURS REVEAL THE LATEST IN ARTIFICIAL INTELLIGENCE
Google looking to hire A.I. experts in China, complementing existing engineering teams in Beijing and Shanghai
Wuzhen, China (May 24, 2017) — Artificial intelligence (A.I.) researchers and practitioners from Google, DeepMind, and Chinese companies and universities presented the latest breakthroughs and applications in the cutting-edge field of machine learning at the “Future of A.I. Forum.” The event took place alongside the weeklong “Future of Go Summit,” during which DeepMind’s AlphaGo A.I. is one-game into a three-game match with the world’s number-one Go player, Ke Jie.
DeepMind’s co-founder and CEO Demis Hassabis discussed the evolution of the company’s software, which progressed from playing Atari games to beating world Go champion Lee Sedol at four out of five matches in March 2016. David Silver, lead researcher for AlphaGo at DeepMind, shared how the latest version of AlphaGo runs on a single TPU (Tensor Processing Unit) machine. This leap in efficiency is possible through a new kind of chip Google revealed at its I/O developer conference last week. The current version of AlphaGo uses ten times less computing power than last year’s and learns much more quickly.
Alphabet’s executive chairman Eric Schmidt said machine learning and A.I. advancements herald “The Age of Intelligence,” adding, “The biggest change in my life has been this explosion in neural networks and deep learning.” During a fireside chat with Zhou Jiangong, CEO of Yicai Media Group, he explained how the new technology increases daily productivity and opens up countless opportunities for businesses, especially in fields of “healthcare, transportation, and government.”
An early Google engineer who joined the company in 1999 and now serves as a Google Senior Fellow, Jeff Dean spoke about leveraging machine learning in consumer products, including Google Photos and Gmail. With the latest updates to Google Photos as an example, machine learning techniques can remove raindrops from photos and add filters to stylize photos in the manner of famous works of art. Dean noted the technology “is making a huge difference and there are many, many more opportunities.”
Jia Li, Google’s head of Cloud and A.I. research and development, talked about the company’s efforts to hire A.I. experts to join its existing Beijing and Shanghai engineering offices. “We are eager to hire the top talent and are currently looking to expand our existing engineering team in Beijing to include A.I. experts,” she said. Li is working to help businesses apply A.I., including at automobile companies and call centers, which use Google’s speech API.
Google’s health research product manager Lily Peng shared how computer tools could help extend the screening of eye diseases to underserved countries, including India, which has a shortage of 127,000 eye doctors serving 1.3 billion people. She noted Google’s machine learning models saw slightly higher accuracy in diagnosing patients than some U.S. board certified ophthalmologists. The potential of this technology extends to other disease detection. For example, Stanford researchers recently used TensorFlow to detect skin cancer from images.
TensorFlow is currently the world’s number-one machine learning repository on GitHub, with usage growing much faster than similar platforms. TensorFlow’s engineering director, Rajat Monga said, “Our goal is to bring machine learning to everyone, and also share research.” Connecterra, a Belgian firm, applies TensorFlow to dairy farming while Australian researchers use the technology to judge the health of sea cow populations.
Google Translate also applies TensorFlow, and Zhifeng Chen, a Google software engineer, explained how machine learning powered recent improvements to the company’s translation platform. With the introduction of neural machine translation techniques, the quality of translations have significantly improved. Combined with “computer vision,” the Google Translation app can now instantly translate signs through a phone’s camera, made possible by a version of TensorFlow that runs right on a mobile device.
Google Arts and Culture Lab program manager Pierre Caessa explored how machine learning allows the world to see six million works of art in the Institute’s vast digital archive with new eyes. He ended with a demonstration of “Portrait Matcher,” a platform which uses a camera to stylize your face in the same manner as those featured in famous artwork. Caessa hopes this and other Arts & Culture experiments will renew global interest in art and make it more accessible to everyone.
DeepMind co-founder and head of applied A.I., Mustafa Suleyman, took to the stage to discuss how A.I. techniques could help address some of our most important socio-economic challenges. “We believe A.I. could be an incredible tool to improve our world,” he noted, “it could help solve some of the pressing and most complex challenges, including those in energy and healthcare.” He highlighted some of DeepMind’s recent projects in this area, including working with Google to “reduce the power required to cool Google data centers by 40 percent."
Other speakers included Professor Gang Chen (Zhejiang University, Deputy Dean of the School of Computer Science, PhD tutor), Professor Zhiqing Liu (Beijing University of Posts and Telecommunications, Head of the Computer and Go Research Center), Zhang Peng (Founder and CEO of GeekPark), Wu Yijian (CEO, Yuanqu Tech), Yili Lin (VP Product, Mobvoi), Yinan Xie (VP, Face++), Bo Li (Director, Smart Driving Development Center, Geely Automobile Research Institute), Yi Wang (Founder of Liulishuo), Zhe Li and Fangzhou Liao (Tsinghua PhD students), and Xiaoming Li (Secretary General of the Wuzhen Think Tank).