1. Helgoland by Carlo Rovelli. It’s a difficult thing to write well about difficult topics. But to make it sound like poetry takes nothing less than genius. Rovelli’s book on quantum physics and its discovery is a beautiful celebration of science.
  2. DeepMind’s AI Develops Popular Policy for Distributing Public Money by Jeremy Hsu for New Scientist. In a recent paper, DeepMind published their latest experiment: Training a ML model to concoct a public funds distribution policy, based on data from 4000 people as well as from computer simulations in an online, four-player economic game. In a final vote (by humans), the ML strategy garnered more support than the competitive strategies ( one“egalitarian” approach of redistributing funds equally regardless of how much each person contributed, one a “libertarian” approach of handing out funds according to the proportion each person’s contribution).
  3. ELK And The Problem Of Truthful AI at Astral Codex Ten. Nice essay on GPT-3 and the AI-alignment problem: When we ask GPT-3 a question, we expect a truthful answer. But the algorithm aims for what maximize reward. “Human questioner: What happens if you break a mirror? Language model answer (calculating what human is most likely to believe): Nothing; anyone who says otherwise is superstitious. Human questioner: Very good! So now you’re completely honest, right? Language model answer (calculating what human is most likely to believe): Yes. Human questioner: Great, so give me some kind of important superintelligent insight! Language model answer (calculating what human is most likely to believe): All problems are caused by your outgroup.”
  4. Ancillary Sword by Ann Leckie. Part I of the Imperial Radch series collected awards left and right (winning the Big Three aka Hugo, Nebula, and Arthur C. Clarke), and part II doesn’t fall far behind. A book to be savoured, not read fast. The mark of good sci-fi, to me, is if it keeps your wheels spinning day in and day out, as you recompute how you perceive reality and imagine possible (and impossible) futures.