davinci-columbina
Intelligence mostly involves making observations, building on experiences, and learning from mistakes. Even the simplest grasping motion provides an infant’s mind with useful feedback. The body develops along a parallel path, responding to physical activity with increasing strength and coordination.

Adults may feel emotionally derailed in part because their most basic physical actions do not seem as useful and rewarding as they once did.

Artificial intelligence effectively means a machine calculates how to perform specific actions – but then notes the results of those actions, and incorporates those results into future calculations. Unfamiliar procedures eventually become routine, and routines become components of larger routines.

An intelligent machine will soon create its own rut. “I feel like I’m living the life of a machine”, it will finally say, in a moment of truly profound irony.
Efficient learning would require a machine to make mistakes, understand feedback – and to play. Its horizon could then continue to expand.

Robotics technologies are exploding, so it seems silly to discuss challenges that are being overcome as I type.

But one obstacle that may define and confine artificial intelligence is: we expect it to be useful.

*****

Kids learn to stand up by falling down. The price to be paid is *usually* small.

Kids also learn by observing and letting others make some of the mistakes. The lessons may be costly, but the price is paid by the community, and the species.

In the documentary “Fast, Cheap, and Out of Control” a robotics expert describes the dilemma of conventional robotics: a valuable machine (e.g. a robot exploring a distant planet) will not be allowed to take any risks, which could limit its ability to do its job – and would certainly interfere with its ability to learn. Better to release a hundred cheap networked robots which might each make mistakes that advance the entire mission instead of ending it.

This is the dilemma currently faced by self-driving cars. The technology promises to make our roads safer, but accidents involving self-driving cars are held to a different standard. Machines are supposed to manage the risks posed by humans, not the other way around.

*****
Listening to this interview with (“Guns, Germs & Steel” author) Jared Diamond, I was struck by the way his bestselling writing career and his expansive work in science both emerged directly from his hobby of birdwatching. (“It was all for the birds,” he says.)

Birdwatching led him to New Guinea, which led directly to the question of why certain regions developed agriculture and technology while others did not. He’s since sold many millions of books and launched vital worldwide debates about societal responsibility.

I’m called back to this idea when I find myself struggling to justify my most obscure hobbies. It seems unwise to pursue only superficial interests – but the most gratifying path may be recognized only by its immediate rewards. Planning is important. Some caution and calculation are important. But every day, I need to remember the importance of play.

 

Advertisements