I went to a panel about machine learning and music in Audio Engineering Society Conference today. The panelists described the products that used machine learning algorithms. One of them was Neutron from Izotope, which used machine learning algorithms like CNN and RNN's to identify a track's instrument. It was cool to see neural networks applied to a musical plugin, but then someone questioned of its usefulness--identifying the instrument of the track can be done without all the fancy machine learning technology. It was the exact question I had, so I was quite impressed by the CTO's response: he acknowledged that it is indeed true that users can identify tracks' instrument without the science and the technology. However, he proposed that this tool to be useful for increasing efficiency in workflow, by allowing the tracks to organize themselves automatically. I thought it was a brilliant delivery of a technical product of its usefulness that may not seem quite convincing immediately. This communication skill is something that I definitely should learn for whatever technical message I have to deliver.