On May 27-28, SYNCED will host the Global Machine Intelligence Summit 2017 (GMIS) at 898 Innospace in Beijing. This is the first GMIS hosted by SYNCED, comprising 5 themed keynote sessions, an intelligent machine expo, an exciting man versus machine contest, and 47 key researchers and industry affiliates.
From deep reinforcement learning, GAN networks to large-scale deep learning, the boundaries of machine learning are constantly being broadened. Talented computers can even outperform top human opponents. GMIS 2017 will host top researchers in the field, who will share their views on future trends in machine learning, including unsupervised learning, voice recognition, noise reduction, visual information processing, transfer learning and cognitive dialogue; while also covering interdisciplinary studies such as linguistics and neuroscience.
We are eight days from the event’s official opening and while attendees have already likely prepared for their favourite keynote speakers, this article previews the Summit’s technical discussions.
List of Keynote Speakers
Jürgen Schmidhuber – godfather of LSTM, Scientific Director of The Swiss AI Lab IDSIA
Stuart Russell – Co-Founder of the Berkeley Artificial Intelligence Research Lab, Professor of Computer Science and Engineering at UC Berkeley
Gary Marcus – Professor of Psychology and Neural Science at New York University, entrepreneur, best-selling writer, Founder and CEO of Geometric Intelligence
Fei-yue Wang – Vice President of Institute of Automation at Chinese Academy of Sciences
Li Deng – Chief Scientist of AI and Partner Research Manager at Microsoft
Dong Yu – Distinguished Scientist and Deputy Director of Tencent AI Lab
Yinyin Liu – Algorithms Engineer at Intel Machine Learning Solutions
Weiying Ma – Vice President at TouTiao AI Lab
Qiang Yang – New Bright Chair Professor of Engineering, Hong Kong University of Science and Technology
Kai Yu – Founder & CEO of Horizon Robotics
Xiaochuan Wang – CEO of Sogou
Martin Müller – Professor and the Associate Chair (Research) in the Department of Computing Science at the University of Alberta, PhD supervisor of David Silver
Day 1: From Unsupervised Learning to Interdisciplinary Inquiries
DAY 1 MORNING: Machine Learning
Father of LSTM Jürgen Schmidhuber: True Artificial Intelligence Will Change Everything
Jürgen Schmidhuber, known as “the father of Long Short Term LSTM” and deep learning neural networks, will deliver an opening speech on the first day. Schmidhuber will guide us through one of the most transformational forces of the 21st century: artificial intelligence, with the premise that AI is about to change most every aspect of human civilization. As a great supporter of artificial general intelligence, Schmidhuber will brief us on how AGI will shape the future of society.
Microsoft Chief AI Scientist Li Deng: Recent Advances in Unsupervised Learning
Microsoft Chief Scientist of AI Li Deng will speak on recent advances in unsupervised learning. Deng argues that traditional unsupervised learning methods such as clustering, GAN, and variational autoencoders (VAE) focus on modelling input data; while his current research introduces a new method called stochastic primal-dual gradient (SPDG), wherein unsupervised learning explores the output structure and techniques proposed for a new cost function. Relevant research has been published on arXiv: An Unsupervised Learning Method Exploiting Sequential Output Statistics.
Tencent AI Lab Deputy Director Dong Yu: Recent Advances in Unsupervised Learning
After exploring the frontiers of unsupervised learning, Tencent AI lab Deputy Director Dong Yu will help familiarize us with advances in speech recognition. Deep learning has made promising progress, however dilemmas like the “cocktail party problem” remain unsolved. Yu will walk us through the historical development of voice recognition technology and its current challenges.
Ohio State University’s Deliang Wang: Speech Denoising Technology with Deep Learning
The afternoon sessions begin with machine learning. Deliang Wang, director of the Department of Perception and Nerve Dynamics from Ohio State University and Chief Scientist at Elephant Sound Technology, will share with us the methodologies of speech denoising technology. Wang is the first scientist to apply deep learning algorithms to voice enhancement. Currently, he is using deep learning to solve the “cocktail party” problem, which will help the computer discern dialogues in a noisy environment.
Horizon Robotics’ Kai Yu: Deep Learning in Autonomous Driving
The founder of Horizon Robotics, Kai Yu, will continue the discussion of machine learning with applications in autonomous driving. Autonomous driving touches upon many aspects of artificial intelligence: computer vision, prediction, and decision making with added requirements to real-computation computation speed. Yu will share with us the technical applications of deep learning to autonomous driving.
Tsinghua University’s Jun Zhu: ZhuSuan: A GPU Library for Bayesian Deep Learning
Tsinghua University’s machine learning group has open-sourced the Abacus (ZhuSuan) software library. Using Abacus, users can utilize the fitting ability of deep learning and multi-GPU for efficient training, generate models in complex environment, make full use of unlabelled data and identify uncertainty using rigorous Bayesian algorithms. In this session, Tsinghua University Associate Professor Zhu Jun will introduce us to the functionalities of Abacus.
DAY 1 AFTERNOON: Interdisciplinary research with machine learning
Beijing Normal University’s Si Wu: Dynamical Principles of Visual Information Processing
Wu Si from the Cognitive Neuroscience and Learning Lab at Beijing Normal University will brief us on recent advances in information processing, explaining why “dynamic information processing is the key to intelligence, and prediction is the key to processing dynamic information”.
Science Advisor on the Movie “Arrival” Jessica Coon: The Linguistics of “Arrival”: Aliens, Fieldwork, and Universal Grammar
Those familiar with the movie “Arrival” will know the broad application of linguistic theory. We are honoured to invite the science adviser of the film, Jessica Coon from the Department of Linguistics at McGill University, to help us interpret linguistic theories in the computer age through the world of aliens, linguistic field trips and universal grammar.
Co-Founder of Cardinal Operations’ Dongdong Ge: In the Era of AI, What Can Operations Research Do?
Dongdong Ge, the Co-Founder and Chief Scientist of Cardinal Operations, Shanghai University of Finance and Technology Research Centre Dean will lead us into the field of operations research. Operations research studies how to make things more efficient (in which “optimization” is often the final goal). Today it is a very important discipline for management, finance, computer, military practices and day to day life. Ge will help us explore the interdisciplinary application of optimizing operations in artificial intelligence.
Roundtable Discussions: Frontier and Interdisciplinary Research with Machine Learning
A roundtable discussion on “Machine Learning Frontiers and Interdisciplinary Research” will conclude the interdisciplinary session. Li Deng, Dong Yu, Si Wu, Science and Technology Director at Bay Labs Johan Mathe, and Professor Zhen Gao from McMaster University will join the discussion to give us more creative insights.
DAY 1 AFTERNOON: YOUNG PIONEERS IN MI
4Paradigm’s Yuqiang Chen: No Free Lunch: Machine Learning Models “A Trade-off Between Wider and Deeper”
Yuqiang Chen, the Co-Founder and Chief Scientist of 4Paradigm, will share his views on the “depth” and “width” of deep learning models: is it better to go wide or deep? Please read about Wide & Deep Learning for Recommender Systems on arXiv to learn more.
UC Berkeley PhD Yi Wu: Value Iteration Network
Joining this session will be Yi Wu, Co-Author of Value Iteration Networks. Wu is currently finishing his Ph.D. at UC Berkley under Stuart Russell and he will be explaining the concept of value iteration, a type of neural network that can learn to design strategies rather than passively follow deployments. VINs can learn to plan, and are suitable for predicting outcomes that involve planning-based reasoning, such as policies for reinforcement learning.
Roundtable Session: Young Pioneers’ Insights on Machine Intelligence
The concluding roundtable discussion, Young Pioneers, will feature Ye Jiang, Ph.D. student at the Department of Computer Science at UC Berkeley; Zachary Lipton, Associate Professor at Carnegie Mellon University; Xiantie Zhang, the founder of Peng Feng technology; Hao Zhou, the CEO and founder of Abundy; and Lu Zhang, partner at Fusion Fund (a.k.a. NewGen Capital).
Day 2: From Transfer Learning to Heuristic Search
In comparison to day one’s intensive technical focus, day two will explore the industrial applications and social values of artificial intelligence. Below are technical highlights of the day.
UC Berkeley’s Stuart Russell: The Past, Present, and Future of Artificial Intelligence
To begin the day, Stuart Russell will present the challenges of artificial intelligence. Russell is the author of “Artificial intelligence: A Modern Approach”, selected as the standard textbook on artificial intelligence by more than 1,300 universities around the world.
4Paradigm’s Qiang Yang: Advances in Transfer Learning
Qiang Wang from the Hong Kong University of Science and Technology, the Co-Founder of 4Paradigm, will guide us through the methodology of transfer learning. Transfer learning is an important research direction in machine learning, as it can help shift applications from one scenario to the other.
University of Alberta’s Martin Müller: Heuristic Search in the Era of Deep Learning
The closing keynote speaker will be Martin Müller, world-renowned computer scientist in chess and the man behind the triumph of AlphaGo in 2016. Professor Müller was the P.h.D mentor of David Silver and Aja Huang — who co-published a paper in the science journal Nature explaining the functioning mechanism behind AlphaGo. Professor Müller will be interpreting for us the importance of heuristic search.
GMIS 2017 brings together top talents from academia and industry. Those interested in the technical research of machine learning will have the chance will see frontier ideas intersect with real-world applications. The technical keynote speakers and panelists will be joined by industry leaders including the CEO of Sogou Xiaochuan Wang, Turing Robot CEO Zhichen Yu, founder of TuPu Tech Mingqiang Li, Baidu Ventures’ Wei Liu, Comet Labs partner Saman Farid, and other representatives who will share their experiences on AI investment and product development.
For more information and an event schedule visit gmis.jiqizhixin.com