Inquiry #6

Short review of “The Vergecast” podcast episode on AI art

This podcast went over a number of things I already new, but also highlighted some of the factors that have led to commercial AI art being available the way it currently is. This podcast explained the process of “diffusion” which is essential for creating the new AI generated images we are working with.  Diffusion is, as I understand it from their description, the process by which an AI generator learns how to break down images into “static” and then reassembles that “static” into coherent images.

The podcast talked about how a company called Stable Diffusion making their work available for free has really driven a lot of the current AI image programs and companies.

They have a discussion on copyright which confirmed what I already knew, but also raised some issues I was unaware of.  The guest talks about how they don’t reveal where they get their training data for their AI generators.  He identifies the question of whether the images produced by the generators might violate fair use laws.  There have apparently been programs set up to pay photographers for using their art to train AI generators, although the guest is skeptical that this will actually work to appease photographers.

They talk about how these companies benefit by obscuring the process through which these images are created, but that it is extremely important for society to understand what exactly is going on.

One thing I found interesting was that the guest was excited and wary about the potential of AI art, but also restrained.  He talked about how if AI art ends up just really being used as clip art it might not be as revolutionary as its being made out to be by some.  He ends on the idea that AI art is just fun and exciting for some people and brings an element of play.