Tae Wan Kim, Tepper School Associate Professor of Business Ethics, speaks about his research on the ethical dangers of artificial intelligence.

Video Transcript:

(Music)

Tae Wan Kim: Now, business ethics means AI ethics, because most business is powered by AI. Without any machine learning technologies, or other similar AI technologies, companies do not make profits.

(Music)

The title of my paper is, "Mimetic Versus Anchored Value Alignment in Artificial Intelligence." In the existing approach, human value just means a set of preferences. Once the preferences are observed and optimized, and then let the machine learn to optimize the preference structure and follow the structure. That's the typical way.

(Music)

Microsoft developed a Twitter bot called Tay, and Tay's algorithm was mimetic. Many Twitterians started tweeting misogynistic and racist and very inappropriate expressions to Tay. And Tay, without any filtering system, just learned everything that the humans said.

(Music)

The other approach is what we call the Anchored Value Alignment. So the value system of artificial intelligence is directly anchored at explicitly expressed ethical principles. We use widely accepted ethical principle within moral philosophy and Lowe and US courts.

(Music)

In the future, probably, there will be companies that sell human values to AIs, because without those companies, AIs will not act in a ethical manner.