The “Parrot Principle” of Artificial Intelligence

I was listening to Rory Sutherland describe some limitations of Artificial Intelligence as it relates to understanding context in human behavior.

Rory says researchers had a parrot look at a ripe banana in color print and on a color screen. The researchers were surprised the parrot didn’t react to the banana the way they’d expected. It’s because parrots see color in a different spectrum than humans. Computers are designed to emit RGB (roy g biv) — the red, green and blue. Our brains “trick” us to see a ripe banana. Not so with the parrot.

He explains reality (price) and sales (behavior) capture a moment in time, or activity over a period of time; but “big data” can’t understand the context in which the human behavior was applied.

Rory worries machine learning may result in drawing the wrong conclusions between reality and behavior in human action by not accounting for the context.

A coworker noted parrots can also be taught to say just about any word in the dictionary, but they have no idea what they mean. If a parrot yells ‘fire’ in a theater, there’s no context. Could prove analogous with NLP.