My recent essay “Reflections on the Millennium Problems” argues that these famous mathematical challenges are not as important as generally imagined both by the public and even by experts. Mathematicians love posing true/false conjectures, the harder the better. My essay suggests that such problems may not capture what’s really important in mathematics.

Now we’ve had the tumultuous past week, in which OpenAI has announced the AI resolution of the Navier-Stokes Millennium Problem and 25 Fields medalists have posted a letter of objection. Finding the yes/no answers to conjectures, they complain, is not what matters in mathematics. Famous mathematics problems are just “landmarks and lighthouses”, they argue, and solving them “is only a tool and proxy for achieving the primary goal of conceptual understanding and insight.”

Of course I agree. Still, one can’t help noting that it was mathematicians who announced the $1m prizes in the first place, and are now blaming the AI industry for taking them so seriously.

With a sprinkle of cynicism perhaps one can make a prediction. As I describe in my essay, the P-vs-NP problem seemed urgently practical in the 1970s but now looks mostly academic — yet experts nowadays tend to suggest that we knew all along it was academic. Maybe we’ll now see a similar effect with the Millennium Prizes? They were presented in 2000 as urgently important to mathematics, but maybe experts will now begin to say we knew all along they were just games.

[13 September 2026]