> This dynamic stems from audio’s technological simplicity compared to other media types. We can see this in how much earlier recorded audio emerged than record

> This dynamic stems from audio’s technological simplicity compared to other media types. We can see this in how much earlier recorded audio emerged than recorded video or video game arcades, and live audio (radio) versus live TV or online gaming. Or how much easier it was to make and record music than shoot and press a film, or design a video game. This is not true for ML. Good realistic speech synthesis is a lot more tricky than images (videos however are another thing entirely).