Super HN

   AI could think in ways we don't understand – evading efforts to keep it aligned (livescience.com) 1 point by Bluestein 1 hour ago