They may do something that doesn't involve as much commuting. They'll be doing something that wasn't being done before. Productivity gains means more work gets done, overall.

The point is that the author's analysis is simple-minded to the point of uselessness. If I were to object to building a hospital in a neighborhood because whenever a hospital is added to a neighborhood it results in a sharp increase in the number of people dying there, you'd rightly call me out for looking at just one specific number while ignoring the overall benefits. It's not like these AIs just pop into existence and spew CO2 like some sort of Captain Planet villain's pollution factory, the AIs are doing something and that has value.