A vivid story is not a likely one, and a new name is not a measurement
A four-and-a-half-million-view video narrates a takeover scenario in the second person, ending with "You have cancer." It cites "the average AI researcher thinks there is a 16% chance of AI causing human extinction." It also says, plainly, "This was just a story. It's not a specific prediction about the future."
In favour- The video says what it is, and the disclaimer is genuine rather than decorative.
- The underlying survey is real: AI Impacts 2023, 1,321 responses on that question.
- The mean is 16.2 per cent; the median is 5. A single heavy tail is carrying the headline number, and "the average researcher" hides that.
- The story form does the rest. The conjunction fallacy, established since Tversky and Kahneman in 1983, is that a detailed scenario feels more probable than the vaguer one containing it, although every added detail can only lower the probability. Vividness is not evidence.
Our takeThe same week, six AI firms signed a voluntary accord with the White House, and an executive order told federal agencies to write "Super Intelligence" instead of "artificial intelligence". A critic's line is the useful one: the companies drafted the principles, they hire the auditor, and the commitment is voluntary. A story is not a probability, and a new name is not a measurement. Ask in both directions what was actually counted.
← All Focus posts