Google's “Superhuman” Game-Playing AI. Is It Truly Intelligent?

December 11, 2018

BIDS Senior Fellow Bin Yu is featured in this NOVA Next article about Google's new Alpha Zero "self-teaching artificial intelligence software," which Google claims "taught itself to play chess, shogi, and Go in a matter of hours" and mastered all three to the extent that it outperforms the current top-ranking AIs in the gameplay arena.

According to the article, however, there is still much discussion among researchers about the realities of artificial intelligence research, the intricacies of critical thinking, and even the nature of the unconscious brain - with no consensus on the true definition of "intelligence." Yu points out that the absolute energy consumption of AlphaZero - in comparison with the energy a human uses to engage in similar activities and to master similar tasks - must also be taken into consideration. According to Yu - who works at the interface of statistics, machine learning, and artificial intelligence - the actual performance of the software may not be as impressive as people think, especially when adding in the person-hours that went into its creation and execution.

Read the full articles here:

Google Says It Built A “Superhuman” Game-Playing AI. Is It Truly Intelligent?
December 6, 2018 | Katherine J. Wu | NOVA Next

A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
December 7, 2018 | Science
David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, Timothy Lillicrap, Karen Simonyan, Demis Hassabis

Featured Fellows

Bin Yu

Statistics, UC Berkeley
Faculty Affiliate