Understanding the ‘Slopocene’: how the failures of AI can reveal its inner workings

What if instead of trying to detect and avoid AI glitches, we deliberately encouraged them instead?