In the 90s, we taught neural networks, and statistical machine learning in AI. Cybernetics (as a term) had been out of fashion for decades by then, if only because it was taken to be how Wikipedia describsed it: the study of "circular causal and feedback mechanisms in biological and social systems" and was the domain of applied mathematics and PDEs, whereas neural networks weren't even recurrent.

No, I learned about ML in another way: working with a CS professor who worked on ML at the time. This was back in the days when autodifferentiation was not well known, so we had to come up with analytic derivatives of our loss function to do gradient descent.

What I learned in Cybernetics was really: how prefix-free codes work (unsurprising given that Huffman was the instructor), how karnaugh maps work, and something about sphere packing being easy in 8 dimensions. Oh, and that Huffman was an arrogant jerk but also an excellent teacher.