Charlie Wezel, who writes the Galaxy Brain newsletter for The Atlantic (it’s really good), has a piece up about AI art, including Midjourney, of which I’ve been posting some renderings of various text and image prompts.
Basically, Wezel explores questions of copyright and fair use, asking:
Then and now, my biggest concern is that the datasets that these tools are trained off of are full of images that have been haphazardly scraped from across the internet, mostly without the artists’ permission. Making matters worse, the companies behind these technologies aren’t forthcoming about what raw materials are powering their models.Charlie Wezel, “What’s Really Behind those AI Art Images?”
He then adds:
Once the tool went public, I watched artists on Twitter share their search findings. Many of them remarked that they had found a few examples of their own work, which had been collected and incorporated into Stable Diffusion’s dataset without their knowledge—possibly because a third party had shared them on a site like Pinterest. Others remarked that there was a wealth of copyrighted material in the dataset, enough to conclude that the AI art–ethics debate will almost certainly get tied up in the legal system at some point.Charlie Wezel, “What’s Really Behind those AI Art Images?”
Definitely questions worth asking. As someone who works in higher education, I am struck by how little this question continues to be asked in regard to plagiarism tools such as Turnitin, whose entire business model is built on unpaid student labor, students whose work is often submitted to the tool without their consent or knowledge. The company has (I believe) been sold more than once for large sums, and none of the “writers” who contributed to the dataset were paid. In fact, they likely were paying an institution (say, in the form of tuition) for the privilege of having their work contribute to a valuable dataset that resulted in no compensation. No one will be surprised that a US court ruled this was “fair use.” It’s really up to the users themselves–those seeking to discover plagiarism–to draw the ethical line themselves.
I am one of those people now. I have very much enjoyed playing with Midjourney. Still, the truth is that I’m not knowledgable enough to know where the images being generated are coming from. Am I actually looking at someone else’s work and the artists themselves are unaware?
One of the first things I tried on Midjourney was a self-portrait (I had just read a Van Gogh biography, and Van Gogh produced a number a great self portraits, one of which was just recently discovered.) Below is what “Chuck Rybak” returned in Midjourney. (For the record, none of those look remotely like me, but that doesn’t mean they aren’t somebody else!)