Spirit Woman, 27th Day of Summer, 525 M.E. (Electra): I have to hand it to Midjourney. In the time since I joined back in October 2023 (version 5.2), they've constantly improved their model from 5.2 to the current 7.0, offering greater realism, more imagination, and more faithful interpretation of prompts. Last week, they added video generation and even in the only four or so days since then, they've tweaked the video model with different download options - "raw," and "optimized" after they noticed that many of the videos were getting overly compressed when uploaded to social media.

Yesterday, I noticed that they also offered the option to download the vids as animated gifs, which work better on this Blogger platform. No more arrows to click to start the video, no more awkward Blogger frames around the video. The gifs function here just like photos, except they move.

Cool.

Look, say what you want about AI - and there's a good healthy debate to be had around its long-term effect, its carbon footprint, and its potential to go rogue. But I like the image generating possibilities of models like Midjourney, and despite what critics say - many of whom never generated a single image - there's more to it, or potentially more to it to be precise, than just typing in some words and getting a picture.

The image above wasn't an instant, random result. It took several iterations and selections from a variety of initial choices, then tweaking one of them to bring out some of the preferred aspects and to suppress some of the others. It wasn't the precise realization of an image I had in my mind, but it wasn't just being handed what the AI decided to come up with. It was a collaboration between a human mind and a virtual mind, and while I wouldn't claim it as "my" work, my fingerprints aren't absent from the final form.

If you go to any one of the many online galleries of AI art, you'll quickly see that some folks consistently produce better, more interesting, and more visually appealing images than others. It's not all the AI, it's also the human collaborators. 

There's an argument to be made that photographers don't "create" their images either - they point and shoot, and the pixels produced in the camera generate the image. Anyone can aim a camera, press a button, and take a shitty photograph. But a skilled photographer carefully frames the subject and composes the image, manipulates the light and exposure, works with the depth of field, etc. The human collaborators in AI art (at least the good ones) work with a different set of controls - skillfully using prompts, culling from several alternatives, enhancing the selection, etc. - to produce their imagery. As we weave and unweave our bodies from day to day, their molecules shuttled to and fro, so does the artist weave and unweave his image.

I'm not claiming I'm an artist. The best pictures I've created were flukes, luck not skill, and even then don't even belong in the same league as the best AI art I've seen on line. But ditto the photographs I've posted, and my writing, and my thinking. 

But despite all of these thing's shortcomings, I don't believe anyone can reasonably say it isn't self-expression.