I never had the reflexively pro or anti AI stance most people seem to have. Its a technology. One misnamed as I have seen little evidence of any intelligence yet being involved in its processes beyond that of a linguistically gifted search engine or calculator. I do believe, however, that much like weapons or vehicles, the question is never the technology itself but rather who makes it and for what purpose.

Under our current monopolistic Silicon Nightmare, the companies pushing this are really just working towards the immiseration of the planet as slop-content mills who drive up electricity bills to grow what could well be the greatest financial speculative bubble of all time. Thus I have never given any of them a single dime. But free trials, using the accounts of other people I know, automatically having access via attached services etc, were once something I experimented with. When this technology was new, it was the correct take to engage with it in order to understand its potential beyond just hearsay. Long time readers know I used to experiment with such images on this very site, before it was commonplace to do so. I have largely stopped doing that, first gradually, and then almost entirely in the past couple of years. My last hurrah was to use pixel art upscaling to illustrate something I wrote (wrote myself, AI writing is so abominably pointless and formulaic for me to even experiment with in any capacity so you can rest assured its not something I have any interest in adopting). It is worth exploring why, aside from not wanting to give money to any of these companies who are engaged in mass theft from actual artists, I stopped doing this.

Leaving behind all the political and economic concerns, something I have been noticing for the last few years is that all of the various generators have become incredibly samey. There may be a plethora of them, but they all create a very specific airbrushed softcore porno look with 90s family photography studio lighting. Even today, now that we have finally entered the realm of people under 60 years old actually having to look twice at some content in order to tell that it is fake, there is a generic look that is especially apparent in the lighting as well as the tendency to make backgrounds repeat details.

The market demand seems to be that people use this technology for meme-making and photo modification, which is fair enough and I have no problem with that. But it has robbed distinctive character from each generative engine. This may not seem like much, since all of this is effectively the creation of fake content for short term amusement, but I think it goes a long way to explaining why I, personally, find it so boring now. Its not just its ubiquity, which is bad, sure, but still not as bad as Corporate Memphis, but rather that I can no longer detect any kind of specific and differentiated nature in any of it. Midjourney, Stable Diffusion, Wombo, etc, they all had distinct looks to them a few years ago.

Midjourney, specifically in its first two versions, was by far my favorite. The reason for this was that no matter what prompt you used, it had this specific directional drive, a style if you will, to move things in the direction of early 20th Century Futurist and Dadaist painting. This made it bad for humorous mashups or portraits but great for surrealistic landscapes and otherworldly visages. In my occasional dabbling I became particularly fond of making what might be described as DAW Books covers for novels never written as channeled through some kind of dream cosmos. Other engines were good at making propaganda posters for nations that never existed. The fact that they looked like something dreamed up by a search engine was not a drawback, but an advantage. By NOT looking like a poor facsimile of an actual artist of photographer, but rather as the fever dream of an internet subconsciousness, I felt this stuff actually had aesthetic value- provided you knew how to prompt it. More importantly, it gave you unexpected almost spontaneous results, as opposed to today’s ultra-literal machine autism which will only surprise you with how boring the results often are.

That seems gone now. Much like with AAA game graphics, the rush for photorealism has sacrificed style and distinction for an endless sameness that is technically more proficient but is actually infinitely less memorable and more likely to date poorly.

As things stand now, I have come to a lower and lower opinion of this technology and where it is taking society at large. Yet I understand that it is here to stay in some form or another and will have some valid uses, not so much on its own as in support of real effort. The only use I still have for it as of now is to occasionally make my own drawings (hand made, done for fun and because its good to have a non screen indoor hobby) have 90s game style pixel art alternative versions of themselves. And this occasional diversion is functionally no different in character from when I used to use Picasa software (RIP to the realist photo editor) to do basically the same thing in the late 2000s-early 2010s.