The changing shape of creativity in this AI era is a fascinating thing for us to consider, especially watching the way that high-schoolers and others interact with the technology and the way it influences their learning and creativity in positive and, of course, negative ways, which are of concern to someone working in the education sector.
We do see, in some sense, a democratization of the capacity to create. The things that high-schoolers can create with the help of some of these AI tools for their images, PowerPoint slides and study notes are truly impressive developments.
At the same time, as you suggest, there are concerns about what this is going to mean for the future of learning and education. Also, what does it mean for the future of creativity? If everybody can create an output that looks like it was made by a leading graphic designer but was in fact created by my 15-year-old, what does that mean for the future of graphic design? These are very real questions we have to confront.
I want to again caution on the relevance of copyright and intellectual property to those questions. I have suggested that we should restrain the expansion of copyright law such that it doesn't protect the outputs that are generated by these machines. I think that's the best way to make sure that the copyright system encourages human authorship and human creativity rather than just incentivizing the mass creation of a glut of AI-generated outputs. The copyright system should remain focused on encouraging and incentivizing humans to create using human skill and judgment and intellectual capacity.
As for how to protect that once it is created, I take you to be saying that it's going to get sucked back into the system and become the training data that others use. I don't think our focus should be on the inputs to these systems. I think when artistic works, outputs or creations are used to train an AI system, that's part of the black box of how these algorithms work.
Our focus should be on protecting human authors against the substitutes in the market, the outputs of AI that might be substantially similar to those works and might compete with them and undercut the market. That's where copyright law should continue to work, as it has always worked. If the outputs are infringing, then the outputs will continue to be infringing, but limiting the capacity of AI to be trained on art, music and literature would be a serious obstacle to its development—and it kind of misses the point.