Algorithms Among Us: The Societal Impacts of Machine Learning
Videos available below. Report here.
Public interest in Machine Learning is mounting as the societal impacts of technologies derived from our community become evident. This symposium aims to turn the attention of Machine Learning researchers to the present and future consequences of our work, particularly in the areas of privacy, military robotics, employment, and liability. These topics now deserve concerted attention to ensure the best interests of those both within and without Machine Learning: the community must engage with public discourse so as not to become the victim of it (as other fields have e.g. genetic engineering). The symposium will bring leaders within academic and industrial Machine Learning together with experts outside the field in order to debate the impacts of our algorithms and the possible responses we might adopt. A particular focus will be paid to technical areas of Machine Learning research that might serve to tackle some of the highlighted issues.
15:10: Erik Brynjolfsson – Presentation on economic issues.
15:30: Ian Kerr – Presentation on legal issues.
15:50: Panel discussion: “Near-term issues” What should we do to make ML/AI as beneficial as possible in the near term? To include economic, legal and data issues (privacy and responsibility), perhaps also military.
Panel: Tom Dietterich, Ian Kerr, Erik Brynjolfsson, Finale Doshi-Velez, Neil Lawrence, Cynthia Dwork. Moderated by Mike Osborne.
16:30: Coffee Break.
17:00: Panel discussion: “Human-level AI – if, how, and when?” Will the quest for human-level AI ever succeed and, if so, when and by what technological path?
Panel: Yann LeCun, Andrew Ng, Gary Marcus, Shane Legg, Tom Dietterich. Moderated by Murray Shanahan.
18:00: Dinner Break, to be followed when we return
19:00: Nick Bostrom – Presentation.
19:20: Percy Liang – Presentation: “On the Elusiveness of a Specification for AI”.
19:40: Panel Discussion: “What if we succeed? Research priorities for beneficial AI” If the quest for human-level AI will eventually succeed, then what research is worth starting today to maximize the probability of a positive outcome?
Panel: Yann LeCun, Andrew Ng, Nick Bostrom, Cynthia Dwork, Percy Liang, Richard Mallah, Shane Legg, Tom Dietterich. Moderated by Adrian Weller.
The organizers would like to extend a special thank you to Richard Mallah for his efforts to help get this symposium set up. The organizers would also like to thank the following sponsors for their generous donations, as well as Victoria Krakovna and Ariel Conn of the Future of Life Institute for their help organizing and designing the