AI with AI
Episode 3.15: The Unreasonable Perplexity of Fat Tails
Andy and Dave discuss an announcement from Exscientia and Sumitomo that they have the first entirely-AI developed drug that is now entering clinical trials. The Director of the Joint Artificial Intelligence Center, Lt. Gen. Jack Shanahan has announced his retirement. Senator Michael Bennet sends a scathing letter to the U.S. Chief Technology Officer on its recent AI Principles for regulation. DARPA's Habitus program seeks to automate the process of revealing and using local information, to enhance stability operations in under-governed regions. And Washington state legislature has at least one facial recognition bill under consideration. Google Research announces its Meena chatbot, which claims is more Sensible and Specific (a new metric that it developed) than the award-winning Mitsuku, though requiring 30 days to compile on 2,048 tensor processing units. Researchers at the Max-Planck Institute for Intelligent Systems and the University of Florence announce a method for fusing deep learning with combinatorics solvers to create a neural network for combinatorial problems. RAND releases a report on Deterrence in the Age of Thinking Machines. Sejnowski pens thoughts on "the unreasonable effectiveness of deep learning in AI." Taleb takes a detailed look at the statistical consequences of fat tail distributions. Maj. Gen. Mick Ryan pens the final (?) part in his trilogy, AugoStrat Awakenings. Fortune publishes a special magazine on the topic of AI. And Andrew Ng and Geoffrey Hinton sit down for a 40-minute chat on deep learning.