I recall seeing a Two Minute Papers episode about an AI built to be curious, or specifically - explore routes in a maze where it couldn't predict the outcome.

I recall seeing a Two Minute Papers episode about an AI built to be curious, or specifically - explore routes in a maze where it couldn't predict the outcome. The easiest way to block it was to present an animated wall. The AI would just stare at it and not go anywhere.