A quick speculation.
I was wondering. Could you take inspiration from ray tracing images and apply that to physics.
The other day I saw the One Pixel Attack Defeats Neural Networks | Two Minute Papers #240. So I wondered. How can this make sense to a network. I mean machine learning networks are also for the universe benefit.
So could a misclassification make sense? Perhaps if light or something else misclassify the output label. So could light change the decisions networks responsible for electron movement?
I mean. Maybe this is the reason for adversarial GAN miss classifications. Which in the light of light makes sense somehow.