How to best contrast AI and Human Intelligence?
Look to human bias, not AI’s inhumanity
Why do journalists writing theoretically fact-based articles go so far off the rails when it comes to AI? Consider Paul Mozur’s comments regarding AlphaGo’s recent defeat of Ke Jie, considered the world’s best Go player, who apparently came down with a case of nerves in the last game:
AlphaGo of course has no heart and feels no nerves, and in the end that may have helped make the difference. Scientists and futurists have pointed to that cold focus as a major reason artificial intelligence may someday take over large numbers of white-collar jobs. Still, that detachment means AlphaGo lacks the human touch required to manage employees, counsel patients or adequately write flowing newspaper features about its own dominance over humans.
AI’s ‘Cold focus’ is a poor characterization of the differences between human and artificial intelligence. It might be better to note human cognitive biases that make people relatively bad at many activities central to work. The so-called ‘human touch’ in management is laughable, considering the rates of employee disengagement, and especially considering the number one reason people leave jobs is ‘bad bosses’.
Mazur also stubs his toe on the complexities of Go:
Tests of the technology in games like Go still mark an early step. Because the strategy options are limited to moves on a board, games like Go are particularly suited to the technology.
This could be misunderstood as asserting that AI can dominate Go because the complexity of Go is a bounded set. But in reality, Go is enormously complex, even though it is based on perfect information (all pieces are visible, there is no bluffing, or hole cards. Note however that AI now regularly wins at poker, which involves bluffing, though.) The number of possible Go games has been compared to the number of atoms in the universe, 10**80.
The author does end on a sensible high note, about humans and AI augmenting each other in freestyle competitions:
Computer scientists say that often the best use of artificial intelligence is not to pit it against humans, but to pair it with them.
To that end, two Go professionals, each teamed with AlphaGo, are scheduled to play against each other on Friday. Mr. Hassabis has said that top amateur Go players, with the help of AlphaGo, can generally manage to beat the software program in a match. In short, a human with a computer is still stronger than a computer.
Welcome to the future of AI-augmented work where we learn to dance with the robots (see 10 skills for the postnormal era).
Originally published at stoweboyd.com.