AI with AI

Andy and Dave discuss the latest in AI news, including the resignation of Samy Bengio from Google Brain, which fired ethicists Gebru in December and Mitchell in February. The Joint AI Center releases its request for proposals on Data Readiness for AI Development (DRAID). DARPA prepares for the quantum age with a program for Quantum Computer Benchmarking. And a separate DARPA program seeks to enable fully homomorphic encryption with its Data Protection in Virtual Environments (DPRIVE) program. A poll from Hyland on digital distrust shows that Americans think that over the next decade, AI has the most potential to cause harm. Amazon introduces the next level of “biometric consent” required for its delivery drivers, which includes an always-on camera observing the driver and gathering other data; drivers will lose their jobs if they do not consent to the monitoring. And Josh Bongard of the University of Vermont and Michael Levin of Tufts University along with other researchers from Wyss and Harvard join together to form the Institute for Computationally Designed Organisms (ICDO), which will focus on “AI-driven designs of new life forms.” In research, Bongard publishes the latest iteration of its mobile living machines, with Xenobots II, using frog cells to create life forms capable of motion, memory, and manipulation of the world around them. Researchers from the universities of Copenhagen, York, and Shanghai use neural cellular automata to grow 3D objects and functional machines within the Minecraft world. And OpenAI Robotics demonstrates the ability for a robotic arm to solve manipulation tasks, including tasks with previously unseen goals and objects, with asymmetric self-play. And the Book / Fun Site of the Week comes from the Special Interest Group on Harry Q. Bovik (SIGBOVIK), which presents “April Fools” research, descriptions of truly absurd, but fascinating, research.