This seems incredibly cynical in light of the current breakthroughs in machine learning, and probabilistic modelling. It really feels to me as though the AI revolution isn't something to dream about anymore, because we're living in it.
I find it difficult to deny the achievement of tangible progress that is implied by, for example, the self-driving car.
UNINFORMED OPINION by a side-watcher: I don't feel there's any AI revolution at all. We simply have slightly more efficient deep learning networks due to better hardware.
Most of the time it feels like the people who are somewhat successful in those branches simply got lucky by randomly mixing elements A, B and C in an unexpected manner and boom -- magic.
In other words, things progress painfully slow and almost always it's due to intuitive shower/sleep revelations than anything else.
The algorithms are improved, and there are more network types, and better understanding of real neurons, but I think you are mostly correct.
I built a program that played tic-tac-toe in -94, using a combination of traversing a problem tree while evaluating the positions using a neural network.
To my understanding, this is basically the same approach used to develop a Go player, only that it took a month training a minimal network to do something useful at all at the time on my small Amiga 500...
Modeling is a new form of pseudo-scientific speculation. ML is closer to reality as long as the training sets are not a dogmatically interpreted noise, like it usually is in finance and other pseudo-sciences.
You could read about a fundamental difference between a properly controlled, replicable scientific experiment and computer simulation according to some abstract/unverified model and why results of such simulations cannot be substituted for experimental results or any form of evidence in my older comments.
I find it difficult to deny the achievement of tangible progress that is implied by, for example, the self-driving car.