AI with AI
Episode 3.25: The COVID Game of Life
In COVID-related AI research, Andy and Dave discuss the joint announcement from Apple and Google on creating a voluntary COVID-19 tracing system that makes use of Bluetooth and anonymous crypto keys. A report in the BMJ screened 27 recent studies describing 31 COVID prediction models and found that all of the studies had a high risk of bias and that the reported performance of the models was probably optimistic. The Allen Institute for AI has updated its COVID-19 Open Research Dataset (CORD) to include “CoViz,” an AI-powered graph visualization tool. And mathematician John Conway, creator of The Game of Life, died at 82 from complications due to COVID-19. In non-COVID AI news, the National Security Commission on AI releases its 1st Quarter recommendations to Congress. Google Brain introduces a deep RL algorithm to the placement optimization problem for computer chip design. And MIT has provided a hub for AI learning for K-12 students. In research, Facebook AI, Oregon State, and Georgia Institute of Technology describe efforts at combining vision and language representation learning, with ViLBERT (vision-and-language BERT), resulting in a single model that can perform multiple tasks, and even leads to improvements on single-task performance. The United Nations Institute for Disarmament Research releases its report on Swarm Robotics. A research paper from Princeton shows a prediction of life outcomes (e.g., the likelihood of layoff, material hardship, GPA, etc) is still really hard. Joseph Blitzstein and Jessica Hwang provide their 2014 edition of Introduction to Probability for free. The Marine Corps University Press freely releases its 2019 Destination Unknown, a collection of short stories written and illustrated by Marines. And the New York Times publishes a Special Report on AI.