2015 : WHAT DO YOU THINK ABOUT MACHINES THAT THINK?

Assistant Professor and Founder, Playful Systems, MIT Media Lab
Tic-Tac-Toe Chicken

 

What force is really in control
The brain of a chicken or binary code
Who knows which way I'll go, Xs or Os

 

"M. Shanghai String Band, 'Tic-Tac-Toe Chicken'"

In the 1980s, New York City's Chinatown had the dense gravity of Chinatown Fair, a video arcade on Mott and Bowery. Beyond the Pac Man and Galaga standups was the one machine you'd never find anywhere else: Tic-Tac-Toe Chicken.

It was the only machine that was partially organic, the only one with a live chicken inside. As best I could ever tell, the chicken could play Tic-Tac-Toe effectively enough to draw any human to a tie. Human opponents would enter their moves with switches, and the chicken would move to the part of the cage that corresponded with the x,y position of the Tic-Tac-Toe grid. An illuminated board displayed both players' moves.

More than once, when I was cutting high school trig, I was standing in front of that chicken, wondering how it worked. There was no obvious positive reinforcement (e.g., grain), so I could only imagine the negative reinforcement of light electrical current running through the "wrong moves" of the cage, routing the chicken to the one point on the grid that could produce a draw.

When I think about thinking machines, I think about that chicken.

Had Chinatown Fair put up a sign advertising a "Tic-Tac-Toe Computer," it would never have competed with high school, let alone Pac Man. It's a well known and banal truth that even a rudimentary computer can understand the game. That's why we were captivated by the chicken.

The magic is in imagining a thinking chicken, much the same way that—in 2015—there's magic in imagining a thinking machine. But if the chicken wasn't "thinking" about Tic-Tac-Toe—but could still play it successfully—why do we say the computer is "thinking" when it was guiding her moves?

It's so tempting, because we have a model of our brain—electricity moving through networks—that is so coincidentally congruent to the models we build with machines. This may or may not prove to be the convenient reality, but either way, what makes it "feel" like thinking is not simply the ability to calculate the answers, but the sense that there's something wet and messy in there, with the imprecision of neurons and feathers.

As opposed to the bounty of precision: it's all about cold calculus. In 2015, it's a perverse state of affairs that it's machines that make mistakes and humans that have to explain them.

We look to the irrational when the rational fails us, and it's the irrational part that reminds us the most of thinking. David Deutsch provides the framework for distinguishing between the answers that machines provide, and the explanations that humans need. And I believe that for the foreseeable future, we will continue to look to biological organisms when we seek explanations. Not just because brains are better at that task, but because it's not even what machines aspire to.

It's dull to lose to a computer, but exciting to lose to a chicken, because somehow we know that the chicken is more similar to us than the electrified grid underneath her feet. For as long as thinking machines lack the limbic presence and imprecision of a chicken, computers will keep doing what they're so good at: providing answers. And so long as life is about more than answers, humans—and yes, even chickens—will stay in the loop.