The Evolutionary AI research group at Sentient has moved to Cognizant Technology Solutions. The group includes several current and past BEACONites, including Risto Miikkulainen, Elliot Meyerson, Jason Liang, and Santiago Gonzalez, and past interns Aditya Rawal and Khaled Talukder. The group has a new website, https://evolution.ml/; the earlier content (announced previously in this blog) is there, including video interviews with 17 academic and industry leaders on “The Future of AI”, as well as the “Evolution is the New Deep Learning” microsite.
Following the idea of expert interviews, the site showcases five new podcasts in the Pulse of AI series (https://evolution.ml/podcasts). In these podcasts, Jason Stoughton discusses topics such as biological vs. computational evolution, trustworthy AI, AutoML, demystifying AI, and open-endedness with Stephanie Forrest, Joydeep Ghosh, Babak Hodjat, Quoc Le, Risto Miikkulainen, Jordan Pollack, and Ken Stanley.
There is also a new site on decision making (https://evolution.ml/esp), featuring research on “Evolutionary Surrogate-assisted Prescription.” The goal is to extend AI from predicting what will happen to prescribing what we should do about it. The idea is to first train a predictor neural network through supervised learning, and then use it as a surrogate to evolve a prescriptor neural network to make good decisions. The site features papers, visualizations, and demos on various game domains (including FlappyBird!) showing how this approach can be sample-efficient, reliable, and safe in sequential decision tasks.
A major new part of the site focuses on COVID-19 (https://evolution.ml/esp/npi): It demonstrates how the same technology can be used to model the potential effects of non-pharmaceutical intervention (NPI) strategies to contain and mitigate the pandemic. The predictor is trained with historical data on the number of cases and the NPIs over time in various countries, i.e. restrictions on schools and workplaces, public events and gatherings, and transportation. A Pareto front of prescriptors is then evolved to discover the best tradeoffs between minimizing cases and restrictions. To illustrate this principle, the site includes an interactive demo: you can explore how, given your preferred tradeoff, the pandemic could be contained and mitigated in different countries.
We invite you to explore the “evolution.ml” site—and perhaps also bring your own expertise in AI to help deal with COVID-19!