Giving a talk on AI progress in Washington this week. What should people know? Stuff I'm highlighting beyond normal = new research showing progress on datasets != general progress (eg @beenwrekt et al CIFAR-10 paper) + @Miles_Brundage et all progress paper, plus robot evaluation.
An important development over the past year is that both researchers and important industry players now appear to be highly aware of algorithmic bias, which was very much off the radar just a year ago.
