AI with AI
Episode 4.15: How Machines Judge Humans
In COVID-related AI news, Andy and Dave discuss research that uses NLP to predict mutations in a virus that would allow it to avoid detection by antibodies. In regular AI news, the US Food and Drug Administration publishes an Action Plan for AI and ML, with more to follow. The White House launches the National AI Initiative Office, which will work with the private sector and academia on AI initiatives. The AI Now institute has launched an effort for “A New AI Lexicon,” in which it invites contributors to provide perspectives and narratives for describing new vocabulary that adequately reflects demands and concerns related to AI technology. And the Federal Reserve is asking for comments about the use of AI/ML in banking, as it considers increasing oversight of the technologies. In research, Michal Kosinski at Stanford University publishes in Nature Reports how facial recognition technology can identify a person’s political orientation (to 72% accuracy); Andy and Dave spend some extra time discussing the challenges and implications behind such applications of facial recognition technology. Researchers at Columbia University demonstrate the ability of an AI observer to “visualize the future plans” of an actor, solely through visual information. The report of the week comes from CNAS on AI and International Stability: Risks and Confidence-Building Measures. The book of the week examines How Humans Judge Machines. And finally, a YouTube documentary from Noclip examines how machine learning plays out in Microsoft’s Flight Simulator.