Machine learning helps tackle the world's "grand challenges"
The third instalment of the Applied Machine Learning Days (AMLD), which ran from January 26-29, sold over 2,000 tickets and welcomed attendees from academia and industry. A diversity of high-profile speakers illuminated the opportunities and challenges of machine learning for health, robotics, risk management, humanitarian aid, and innovation itself.
Hosted by EPFL at the SwissTech Convention Center, AMLD 2019 kicked off with two days of skill-building workshops, followed by the main conference of tracks and talks, plus a poster session and job fair.
The main conference was opened on January 28th by AMLD organizers Marcel Salathé professor and academique director of the EPFL Extension School, and School of Computer and Communication Sciences (IC) professors Robert West and Martin Jaggi. West announced in the introduction that the AMLD event is looking to expand.
“AMLD is planning to go global. We are thinking of AMLD New York, Rio, Tokyo, Kinshasa, St Petersburg, you name it. The mothership will always be in Lausanne, but we are aiming to spread the love,” West said, inviting attendees to apply to bring the conference to their home institutions.
The future of AI through Google’s eyes
Jeff Dean, Senior Vice President of Google AI and Health, delivered the opening keynote with an overview of the impact that machine learning – the study of algorithms that are driven by and ‘learn from’ data rather than code alone – is having on some of the greatest engineering challenges of the 21st century. His talk focused on the subfield of deep learning, which uses artificial neural networks inspired by the structure of the human brain.
Examples of the impacts of these technologies included self-driving cars for improving urban infrastructure, and deep neural networks for screening medical images for disease risk. Machine learning is also being used to improve the process of innovation itself, for example by transforming the way powerful computers are designed.
“Deep neural nets and machine learning are really helping make headway in some of the world’s grand challenges,” Dean said.
He also emphasized how these applications have been made possible by the extremely rapid progress of machine learning in the past five years, with the latest computer models being able to correctly label a never-before-seen image with a 3% error rate, compared to 26% in 2011.
“That dramatic improvement really has fundamental consequences for lots and lots of things. Computers didn’t use to be able to see; now, they can see.”
Industry, insurance, and the global good
Michael Baeriswyl of Swisscom and Costas Bekas of IBM Research offered insights into how companies and industries can best harness advances in machine learning and artificial intelligence (AI). Meanwhile, Jeffrey Bohn of Swiss reinsurance company SwissRe also talked about how AI is transforming the risk assessment and risk management industries. Patrick Barbey of Swiss innovation promotion agency Innovaud also talked about how the region’s density of startup accelerators, higher education institutions, and innovation parks makes it a prime candidate for an “AI Valley”, to accompany the so-called Drone and Health Valleys.
From the humanitarian sector, Charlotte Lindsey-Curtet of the International Committee for the Red Cross (ICRC) and Ramesh Krishnamurthy of the World Health Organization (WHO) related to Marcel Salathé in a panel discussion why their respective organizations are investing in machine learning. In addition to having great potential for digital health and public health surveillance, as well as responding to victims of armed conflict, there are also risks that need to be understood in terms of the use of AI in autonomous weaponry, Lindsey-Curtet and Krishnamurthy emphasized.
Robert West also hosted a panel discussion with Yuanchun Shi of Tsinghua University, Li Pu of Segway Robotics and Adam Knight of Oxford University on how AI and machine learning technologies are being developed in China.