Why the skepticism? Even at ECMWF, they say the AIFS beats the ECMWF in the majority of cases. They do feel that the physics-based models are better at handling the extreme events, which makes sense, in that extreme events have little historical precedent. It would seem that a combination of the two is the best approach now, but I would look for AI to become the go-to soon. Then comes quantum computing, ha ha!