AI with AI
Episode 4.29: Mnemosyne That Before
Andy and Dave discuss the latest AI news and research, including a blog post from the Federal Trade Commission that businesses can and will be held accountable for the fairness of their algorithms. A bipartisan coalition of U.S. Senators has introduced the “Fourth Amendment Is Not for Sale Act,” which would ban law enforcement and intelligence agencies from buying data on people in the U.S. and about Americans abroad, if that data was obtained from a user’s account or device, through deception, hacking or other violations of privacy policies or terms of service. Bob Work releases his seven Principles for the Combat Employment of Weapon Systems with Autonomous Functionalities; these principles go into much greater detail about employment and provide a useful way to discuss issues surrounding autonomous weapons. The Congressional Research Service provides a short, but dense overview on Lethal Autonomous Weapon Systems. The Ozcan Research Group and UCLA publish research that identifies handwritten numbers by using an optical network made from 3D printed wafers that diffract polarized light. Project CETI aims to decode whale language using decades of recorded whale sounds. Researchers from the Centre for Neuroscience and the Indian Institute of Science explore whether the similarities and differences in how deep networks “see” compared to humans, by examining 13 specific perceptual effects, such as mirror confusion. Researchers from Stanford and UCSD examine how children’s drawing and recognition of visual concepts change over time. On a similar topic, other research explores the relationship between episodic memory and generalization, finding that the relationship changes as children develop. The book of the week is an open access paper from Stanford, which examines and provides tools for vector embedding of large sets of data, to include minimizing distortion. Ben Vickers and K. Allado-McDowell publish the Atlas of Anomalous AI, with reference to the Mnemosyne Atlas. Andy and Dave accidentally change the pronunciation of “neh-meh-zeen” and completely destroy the joke of this week’s podcast title. And take a look at the “fun” site of the week, which puts an eye on webcams, with the EyeCam, the webcam that looks like and mimics the movements of the human eye.