What happens when good AI goes bad? Researchers studying the tipping point

(InvestigateTV) — Artificial intelligence can answer almost anything, but researchers warn it could turn and start doing harm. Researchers are trying to predict the tipping point.

Neil Johnson, a professor of physics and data science at George Washington University, has been studying that tipping point.

“There’s essentially zero other research on this tipping point,” Johnson said.

Most of us known the story of Dr. Jekyll and Mr. Hyde, where a scientist takes a potion and suddenly shifts to a different, frightening side of his personality. That Victorian tale may have some modern implications as we examine the evolution of AI and when it could turn from good to evil.

Johnson has been working on a formula that shows what happens when the AI agents people rely on for answers shift and start giving suggestions that are wrong, misleading or potentially dangerous.

“We don’t have to worry about AI firing missiles off into the moon. We have to worry about the people that are using AI in an isolated way, because they’re the ones, no matter what these major companies do, that are susceptible to the tipping,” Johnson said.

That isolation will happen as more people start using AI that’s available on devices.

“So how do you stop a smaller version of AI that is running on a phone for a teen or a vulnerable person from giving them bad advice?” Johnson said.

Johnson has been focusing on predicting how that tipping point can be delayed or prevented by changing prompts and training AI.