<div dir="ltr"><div dir="ltr"><div>Two articles from MIT Tech Review.</div><div><br></div><div><a href="https://www.technologyreview.com/2022/07/19/1056176/sonys-racing-ai-destroyed-its-human-competitors-by-being-nice-and-fast/" target="_blank">https://www.technologyreview.com/2022/07/19/1056176/sonys-racing-ai-destroyed-its-human-competitors-by-being-nice-and-fast/</a><br></div><div><br></div><div>Training AI drivers for Gran Turismo racing, it turns out that they can learn to physically drive faster than people, but they're too aggressive to win head to head races because they drive the competition off the road.  So you need to train them to observe the norms of the competition, by including penalties for crashes, bumps, cut-offs, etc, into the training.  They still drive faster than people, and the way they drive is a bit disturbing to watch.</div><div><br></div><div>[So if you were training AI drivers for political races, would the norms come from established law or where the voters could be persuaded to mark their polls?]</div><div><br></div><div><a href="https://www.technologyreview.com/2022/07/20/1056219/weed-influencer-and-scientist-feud-over-cannabis-hyperemesis-syndrome/">https://www.technologyreview.com/2022/07/20/1056219/weed-influencer-and-scientist-feud-over-cannabis-hyperemesis-syndrome/</a><br></div><div><br></div><div>The scientist and the instagram influencer attempt to study the genetic causes of CHS (cannabinoid hyperemesis syndrome) and the influencer blows up the study when she declares that the scientist is a shill on her channel.  Ah the joys of decision making with uncertainty and cognitive bias.</div><div><br></div><div>-- rec --</div></div></div>