Early 90s, I had to teach a class to psych undergrads, and I got neural networks. The book had already been decided: Rumelhart and Mclelland's PDP. There was software, we had a computer lab, everything was set.
For the first class, I had prepared a bit of an overview, starting with the perceptron, and then showed Minksy and Papert's proof that it could only learn a linearly separable membership function. So I had these formulas on the blackboard, ending with something like "for every delta > 0 there's an epsilon such that ..." etc. Then before the break, I ask if anybody has questions. One person raises their hand and says: What's a vector?
That was a bit of shock. I had completely not understood the knowledge level of my students. Not that I ever became a good teacher, but that was an eye-opener.
For the first class, I had prepared a bit of an overview, starting with the perceptron, and then showed Minksy and Papert's proof that it could only learn a linearly separable membership function. So I had these formulas on the blackboard, ending with something like "for every delta > 0 there's an epsilon such that ..." etc. Then before the break, I ask if anybody has questions. One person raises their hand and says: What's a vector?
That was a bit of shock. I had completely not understood the knowledge level of my students. Not that I ever became a good teacher, but that was an eye-opener.