This week, I’ve been doing a lot of thinking about the implications of artificial intelligence. One of the most fragile parts of the puzzle is the training data. We already know that you can see if your images were used to train the datasets and that a lot of the training datasets out there are . . . spurious at best. Some startups are trying to build datasets trained exclusively on licensed data, and human artists are pretty grumpy when big-name studios use AI to generate art.
