In the wake of the Conference on Innovative Data Systems Research (CIDR), several of its high-profile participants will speak at a special public meetup ADS Meets CIDR. Turing Award winner Michael Stonebraker will talk about data integration, and Christopher Ré about weak supervision and the Snorkel system. The lineup is completed by Hinda Haned, who will speak about explainable AI.
‘Most of us are working on the wrong problem.’ With this proposition, Michael Stonebraker expresses in no uncertain terms how he feels about the way data science systems research has developed in recent years. Stonebraker, professor at MIT, won the Turing Award in 2014 for his pioneering work in the field of database systems. He is founder of more than ten data-related startup companies, among which TamR, a company focused on automatic data cleaning and data integration, where he serves as CTO.
High-quality data for high-quality machine learning
The second speaker at ADS Meets CIDR is the also quite famous and entrepreneurial Stanford Computer Science professor Christopher Ré, who is affiliated with the Statistical Machine Learning Group and Stanford AI Lab. To get good machine learning results, a data scientist needs to possess a lot of high-quality training data. This is a dilemma that can make the promise of machine learning sound as useless as ‘If you want to get rich, get a lot of money’, which is in the title of Ré’s talk.
Ré will discuss his work on Snorkel, a highly recognized system for rapidly creating training data using weak supervision. Chris Ré founded a number of companies based on his research, the last one of which was acquired by Apple in 2017, hence his double affiliation (Stanford and Apple).
Bringing explainable AI into practice
The third and last speaker of ADS Meets CIDR is Hinda Haned, Chief Data Scientist at Ahold Delhaize and professor by special appointment at the University of Amsterdam. Haned will discuss some of the challenges of bringing explainable AI into practice. Explainability should be thought of as a process rather than a product, according to Haned. Providing explanations about how a machine learning model produced a particular outcome can help enhance users’ trust and their willingness to adopt the model for high-stake applications. In recent years, we have seen a surge in research on explaining AI-powered systems. However, Haned argues, very little in this body of work evaluates the usefulness of the provided explanations from a practical human-centered perspective.
The ADS Meets CIDR meetup, organized by Amsterdam Data Science (ADS), is chaired by CWI researcher and VU professor Peter Boncz, and takes place on Wednesday 15 January 2020 at 15:00 in the Mövenpick Hotel Amsterdam. Registration is free, but required via the ADS Meetup page.
