What is an AI anyway? A vendor's answer, and a classroom's

Mustafa Suleyman, TED

The talk is from April 2024 and has been played about 3.7 million times. Mustafa Suleyman, who runs Microsoft's AI division, opens with his nephew asking over breakfast what an AI is, and admits that his stock answers, a tool like printing or steam, sounded like a policy seminar. The answer he offers instead is that AI is best understood as "something like a new digital species," and he predicts personal AIs that will be tutor, doctor, lawyer and coach in every pocket, and "companions, confidants, colleagues, friends" besides.

He says twice that the species idea is an analogy and not a literal description, and the analogy is the part a faculty can argue about, which is why the talk is worth its twenty-two minutes. A student who calls the chatbot "him" and says he built a relationship with it, as one Dayton sophomore told NPR on September 28, is describing the companion Suleyman predicted. Wichita's AI specialist teaches elementary students the opposite, that the thing is software and is not a friend. New York City has barred companion chatbots in every grade, and Florida's pending rule bars AI built to meet a student's social or emotional needs. Two and a half years after the talk, the companion half of the prediction is what school rules are being written against. The tutor half has not yet been shown to work in an American classroom. Stanford's review found no rigorous study of student AI use in U.S. K-12.

We would show a class the opening and stop at the nephew's question, a little past the two-minute mark. Students write their own answer before they hear his. Then play his, and ask what changes if a program that talks like a friend is called a species, and what changes if it is called software. Students can use Suleyman's own words to explain where they agree with him and where they do not.

Our guide to what AI literacy is sets out what a graduate should be able to say about a system like this one.