The "What"

So I do a lot of talking about AI, it's a very popular technology right now, so there's a lot going on with it. It's permeated every bit of our lives in some way and, while it's certainly an interesting technology, it's got a lot of people on edge.

One of the big problems that AI causes is in AI generated images. Because people can generate almost any image, it's very easy to fake events. It's to the point where even a photo of Mitch McConnell in the hospital is scrutinized for potential AI tampering (it wasn't). This has lead people to speak out against AI imagery and point it out when it's not labelled as such. So with that I thought I'd go and do an interesting little social experiment: I would use my own AI imagery to see whether people would actually point it out as such.

Disclaimer: This is not a scientific study, nor is it evidence for anything. This was purely anecdotal and for the sake of entertainment and a little bit of social anthropology. Do not interpret anything in this blog post written by some dude with too much time on their hands as anything other than pure curiosity.

The Experiment

Preparation

So, a couple of months ago I swapped from a profile picture that was my Memoji, a cartoon persona you can make on Apple devices, to an AI-generated image of myself. I didn't just want it to guess, so I gave it an actual image of myself on which it could base its own output. The photo I chose—my current profile picture—was one that I actually took myself and then passed to the model as a reference.

Execution

Once I had the image generated, I swapped my profile picture and said nothing about it. Over the next couple of months, pretty much nothing happened. Nobody said anything about the profile picture and I just went about my day. After a while, I swapped out the original image for something a little more obviously AI-generated. Instead of basing it on a real picture of me, I fed the model the original AI-generated image and explicitly prompted it to make it look more cartoonish. I swapped that one in and went about my business.

Criteria for Noticing The Image

Now, I'm not going to say that people didn't notice my AI profile picture at all, the criteria here is "meaningfully notice" in a way that's actually measurable. Effectively, I'm just waiting around for someone to actually comment on it. I can't read minds, so someone's got to be outward about it.

When People Started to Call It Out

For the longest time, nobody actually went about calling it out. Again, most people went about their days posting and interacting with my content online. Nobody was outwardly bothered by it. That is, until people got mad at things I said.

The particular post that actually got someone to comment on my profile picture was this one right here, posted on The Fourth of July.

While I'm not quite sure what about this post prompted this particular response, it was the first time someone actually made a comment about the post, so it was finally a turning point in my little social experiment.

One clarification that I'd like to make is that I wasn't hiding the fact that the image was AI-generated, I was simply not saying anything unless someone explicitly brought it up.

Here, I was the first person to suggest that it was because of some kind of AI detection.

The End of the Experiment

The conclusion of the actual experiment actually came not long after the above interaction, interestingly enough. A while back, GitHub had posted a link to a blog written by a neurodivergent C-suite exec and their experience using AI as an assistive tool in getting their job done.

My response was this:

That one seemed to get people mad (and also was proved right by most of their responses), but at some point, this post had resurfaced earlier this week and got a lot more interaction. People calling out the profile picture as if that somehow invalidated anything I had to say.

Note: nothing I said actually was defending GitHub

Thoughts

So I find it particularly interesting that people didn't seem to really even care that the images were AI generated until they wanted to use it as a arguing point. I don't really have an actual conclusion about any of this, but I feel like there is something to be said that nobody really seemed to comment on the images until this point.

Anyway, I've put my profile picture back to the actual picture of me and I figured I'd write this blog post about it. I'm not trying to draw a conclusion about anti-AI sentiment here, I'm really just chronicling a particular experience I had with AI image generation tools. Make of that what you will.