4 Reasons We May Not See Colluding Robots Anytime Soon

Law360, New York (October 3, 2017, 1:00 PM EDT) -- Many concerns have been raised in the antitrust community about algorithmic collusions. I summarized some of the views and concerns in a recent article titled “When Machines Learn to Collude.”[1] In that article, drawing lessons from a recent artificial intelligence research, I highlighted several observations, including the lack of empirical evidence that the singular goal of profit maximization would lead to collusion by machines. Instead, to design successful colluding algorithms, learning to collude likely has to be an explicit design feature. If a firm adopts such...
To view the full article, register now.