The "nightmare" seems to be that the LHC only confirms the standard model without giving clues on how to extend physics further. I've got a hunch that a future breakthrough may come from AI, that combining general relativity and quantum mechanics is a bit much for unaided human brains and deep mind style computer systems may do better at spin-2 stuff in odd numbers of dimensions than us lot.
Suppose that you set a computer to work and it learns a fantastically accurate model that no human is able to understand. Do you then declare victory and move on?
I think that depends on if it's garbage (noise) or an actual understanding (signal).
'AI' solutions don't always solve the problem you're trying to train them on, sometimes they solve the problem of giving you the expected answer from the samples you give them.
I'm guessing it would be easier to understand the stuff than come up with it. If it did come up with a model that predicted something like particle weights we could then try to figure what it did.
