Yes, but there’s substantial overlap.

Suppose I’m better than my sister at composing classical music, but she’s better at writing novels. There’s no obvious way to judge which of us is “smarter” than the other, since music and novel-writing are just different skills. So how is it any more meaningful to say that an AI is “smarter” than a human?

Our response is: If I’m better at one thing and my sister is better at a different thing, then it may be hard to make meaningful comparisons. On the other hand, if I’m better at one thing and my sister is better at two thousand things, then it starts to seem a little silly to insist that we’re on a level footing — or to insist that there’s nothing we can say about the footing we’re on.

If Anyone Builds It, Everyone Dies is a book about the likely practical impact of future progress in AI. To speak meaningfully about that impact, we don’t need to be able to compare ChatGPT, humans, and fruit flies and say precisely what “intelligence level” these three very different systems are at. We only need to see that AIs are becoming better and better at an ever-wider range of skills, and that eventually they will surpass humans on skills of immense practical importance.