I am excited to see something like this, but this website would inspire way more confidence if there was an "about us" page describing who/what/why.
On the other hand, some of the courses are written by very prominent people in the Julia community, so there is some minimal amount of legitimacy already present.
Yes, we'll put it up shortly. It is basically put together by Julia Computing, and we expect to recruit more content now that we have a basic set of materials.
The creator of Knet is a professor and he teaches a class using Knet, so he must have tons of material available. If you look at the Knet docs compared to the Flux docs, the Knet docs are much more educational and explains things a lot better.
The first-class auto-differentiation seems to me like the special bit in Julia for machine learning (not only neural networks). Flux/Zygote is only one of the tools for that (but it is indeed what I view as most promising at the moment). Do check out cassette as well. However, for the restricted set of autodiff supported by libraries like Tensorflow, basically any of the Julia tools are good enough.
On the other hand, some of the courses are written by very prominent people in the Julia community, so there is some minimal amount of legitimacy already present.