Machine-Learning Maestro Michael Jordan on the Delusions of Big Data and Other Huge Engineering Efforts - IEEE Spectrum spectrum.ieee.org/artificial-int…
A very interesting read, with a healthy dose of anti-hype skepticism, particularly about the dangers of torturing data until it confesses to something... 
This is just the multiple testing problem, magnified by the huge dimensionality of modern datasets. (And the hubris in thinking you can correct for it)
He also has some interesting quotes about how "neural" neural networks are - I'd be interested to know what AI/comp neuro folks think of them:
I would have thought that the Bayesian prediction-error-up-model-updating-down framework is doing something like back propagation. Is that not how people see it?
Similarly, on how learning is instantiated, my feeling is that it is indeed by changing synaptic weights:
Not that I think neural networks are actually brain-like - far from it. But on those two points, it seems there are ~corresponding neural mechanisms
