The way I see it. If you have something like high velocity noise particles. Then some machine learning model() over that. Could perhaps change decisions over time.
That is. Change the ?exact time for an event in the process. By changing time like described by time dilation examples.
Basically there exist a ‘why’ time dilation would exist.
So my guess is that there exist a model(fast_noise_particles) == A.I observer In fusion. The noise particles is here I think the plasma.