Gigaom AI Minute – April 17

Transcript

We each have hundreds of cognitive biases. This is where the brain arrives at potentially incorrect answers because it has certain built-in preferences. My favorite example is the “rhyme-as-reason effect.” Because of it, a statement is regarded as being more accurate if it rhymes. And everyone born before 1980 remembers Johnnie Cochran’s oft-repeated statement about the glove in the O. J. Simpson trial: “If it doesn’t fit, you must acquit.” The jury agreed.

Perhaps these biases are not bugs in our brains’ source code but serve very real purposes. Maybe they only look irrational from a certain point of view. I have often thought that if entrepreneurs knew their real chances for success, vastly fewer enterprises would be undertaken. But because of an optimism bias in such matters, lots of people think, “Sure, most things fail, but mine won’t!” and go on to start companies.

This actually might be the optimal choice from a societal point of view. Would an AI develop its own cognitive biases? Or, would it inherit ours? Or neither? It would be delightful, wouldn’t it, if the way we keep ahead of the computers is that we make individual irrational decisions they would never make?