AlphaGo and the Limits of Machine Intuition

HBR STAFF With the lopsided 4-1 rout by Google’s AlphaGo over Go grandmaster Lee Sedol, the easy takeaway is that artificial intelligence (AI) has achieved another milestone against humans, raising the specter that machines may eventually replace people, even managers. But by winning even in such convincing fashion, AlphaGo has revealed that AI still has a number of shortcomings, particularly when it comes to machine-made intuition. Google acquired DeepMind, the developer of AlphaGo, in 2014, in a $500 million bid to expand its burgeoning AI portfolio. AlphaGo’s deep-learning algorithm allows both a “policy network” and a “value network” to store not only millions of past games played by the masters but also those played against tweaked versions of itself. The naming of the two networks is managerial-sounding and is aimed at promoting efficiency,…


Link to Full Article: AlphaGo and the Limits of Machine Intuition