Comment by jean-

> When the state-of-the-art is 97% on a task, there's only so much room for improvement.

When models commonly achieve 97% on a task, it means it's time to define a harder task, as it's long stopped providing any useful signal.