The generator for a simulation may not be able to predict what the real Earth would look like correctly given infinite Earth-like planets, but that doesn't mean

The generator for a simulation may not be able to predict what the real Earth would look like correctly given infinite Earth-like planets, but that doesn't mean that this algorithm wouldn't be able to correctly deduce some input parameter for the real Earth after training in the sim (like the number of days in a year based on a single minute of simulated video from the surface). It's a bit of a stretch to call it proof, but the article says they spent half a year of work on making sure it's not just an artifact of the simulation they're basing the prediction on. I wouldn't be surprised if this did actually work, if we could ever collect equivalent real-world data.