The Dark Side of Competition in AI

by Liv Boeree from the TED STAGE

(Courtesy of TED)

At the Technoskeptic, we tend to be more than a little leery of all the techno-utopianism purveyed by TED talks. The trash-heap of “next-great-things” promised at TED talks-but which didn’t work-is impressively high.

To toot our own horn the precisely appropriate amount, our own Kentaro Toyama was appropriately skeptical about whether the KhanMigo ChatBot was going to be a revolutionary children’s tutor, as promised on a TED stage almost exactly a year ago.

Over and above KhanMigo’s failure to address students’ motivation, Toyama’s central complaint, we’ve come to learn recently, courtesy of the Wall Street Journal, that the KhanMigo math tutor has some pretty basic issues. It sometimes can’t add, and when it makes a math mistake, it refuses to correct errors. Suddenly, not sounding like a great math tutor? Or a great TED talk?

Still, we cannot throw all TED babies out with all the dirty TED bathwater, and sometimes truths are uttered from the TED stage.

Live Boeree’s explanation about how the race to develop AI is falling into “Moloch’s Trap” seems one of those times.

Of course, the unspoken premise of Boeree’s talk is that IF AI companies avoid Moloch’s Trap, a pretty big IF, the technology they unleash on the world will then be worth the massive investment resources invested, and won’t inflict more downsides than upsides through, to pick the most likely example, destabilizing levels of mass unemployment.

The premise that even the best version of AI will be a net positive for society remains a very long way from being proven.

Enjoy.