I want to know if we can tell the difference about which image came first. Look at the two below. The first is hand-drawn by yours truly. The second either comes from OpenAI’s GenAI image generator or I stylized it myself. Did I draw mine first and give it to AI to stylize? Or did I prompt for a specific image and then do a hand-drawn version from it? Or did I draw mine and then also digitally update it by myself?

How can you tell? Are you sure the second image went through AI image generation or am I making that up to be a “gotcha they were both done by humans” kind of ploy?

What’s the consequence of being wrong in your assessment? What if you said that I couldn’t have drawn that without the AI version? What if you said, that’s definitely human first and then sent to AI? What if you said, but why did you send it to AI, can you not color it yourself? What if both are actually done by me since the coloring and straightening out the shapes really isn’t that hard via GIMP, Photoshop, or other “hand” authored digital image editing tools.

GenAI image generation is a polarizing technology in the art community with vehement opposition of its use by artists and allies, but with staunch support for the technology from corporate brass. Don’t believe me? Just search GenAI in game dev communities on social media platforms and then talk to artists, privately, in-person for a non-social media take.

For me this question is interesting because it can be hard to tell the origin of an image. We’re already accusing people of AI slop in their writing by trying to read the tea leaves when the author uses the word “delve” or “honestly” or if they use em-dashes. Some folks are purposefully adding in typos to “prove” that their work was 100% human produced.

AI detection tools are notoriously poor at identifying if a passage of text is AI written or human written. The consequences of using these tools in education is damning because the false positive rate is too high relative to the consequence imposed (expulsion, failure of class, etc.).

For image generation there are watermarks that can give it away as well as artifacts in the lower bits or when analyzing the image’s color spectrum. So we might be able to more accurately state if an image is AI generated or not. But what about order of events? Where did the inspiration come from? Is it bad to have prompted an image, then hand-drawn it? Is it bad to have hand-drawn it first and then prompted for a “faithful” clean up?

Of course, one camp will always lean on “any AI is bad AI” and so be it that’s one end of the spectrum. The companies offering these tools are on the other end “our AI is good AI”. But in between there is a vast gray with arguments against pointing to damage to environment, cognitive abilities, etc. as well as arguments for pointing to expansion of skill, increase in productivity, enablement of smaller (perhaps solo) teams to do more and cut through bureaucratic red-tape at their much larger orgs where good ideas merrily die.

Back to the images, which came first? The hand-drawn one or the digital art one? And is the digital art one really GenAI?