This is likely to be a rambly kind of a post. I typed the blog title first, but, as Hector Barbosa might say, they’re more of what you’d call guidelines than actual titles. Fair warning – as at least one reader will point out, this may be some more AI bellybutton gazing.1
I basically stopped using Creative Commons pictures in my blog posts and websites a few years ago when I learned of some CopyLeft troll nonsense litigation. I did my best when using these pictures to provide proper attribution and even wrote a WordPress plugin to help others do the same.2 Learning that a picture from Flickr or posted for free use under Creative Commons could lead to abusive litigation and extortion was so disheartening that I pulled a lot of pictures from my various blogs and either left them with no images, broken images, or if they were very important to the post/page making something to go there. From that point, I generally trended towards using/making memes to add visual interest to the blog posts.
I was reminded of this after reading a well written post on Mastodon about various alternatives to using AI generated images in blog posts.3 While I don’t think these are all necessarily practical when it comes to finding an appropriate and aesthetically pleasing image to go along with a blog post, I thought this was a worthwhile and thought provoking suggestion.
I do like adding pictures because it adds a little bit of visual interest to what would otherwise be a wall of text. Many times, I will add a screenshot from OpenSCAD or a photo of a sketch / sketchbook page. And now, a short digression…
Someone far more pithy than I comment recently that twenty years ago regular people were sued for posting song lyrics, sharing MP3’s of music, and low quality movies, while these days people are becoming AI millionaires by shoveling the entire internet and the collected written works of the human race into machines.
…What should people opposed to AI slop do, then? Do we stop driving because of license plate scanners and stop going out in public because some asshole will be wearing Google Glass Goggles or Facebook Meta Glasses and feeding our information into facial recognition software? 45 I don’t share photos of myself or family on social media, but should we not store our photos in the cloud and avoid fun community events at the local library? Even writing this post, I’m unwillingly contributing the corpus that will be fed into the next version of an LLM.
I feel damned either way. If I don’t use any LLM, they will consume my writings and original pictures, use my original pictures to track me across the internet, sell me things or sell me to other things. If I do use an LLM it will do all of that plus train on my interactions.
What is the most defiant thing to do when this blog post is almost certainly going to be read by more web crawlers, spam robots, and LLM’s than humans? Should I only post new pictures to add visual interest to my posts at the expense of embedding even more meta data into each post? Is it more subversive to use an AI to generate a fitting copyright-free but soulless image that adds little new information to the algorithm’s endless hunger?
I don’t have answers to any of this.
As I was working yesterday another thought occurred to me. Where is the line between “software” and “LLM/AI”? I had just run optical character recognition on a large document to help figure out what was in it. All OCR is just pattern recognition, making trained “black box” guesses at what’s a given letter/word/pattern. I’ve never liked the auto-suggest “feature” of my phone’s texting app or gmail, but I like the auto-complete feature of LibreOffice when I start to type a date. I don’t mind when my text editor suggests the name of previously used variables and functions – it’s actually quite helpful.
Why am I comfortable with an heuristic’s OCR/autocompleting dates/autocompleting variable/function names, but I hate auto-suggesting messages? Perhaps because it’s no something I can turn off. I just know I find these kinds of “helpful suggestions” to be far more intrusive. Maybe I don’t mind these helpers in documents – but I don’t want them stepping the communications between myself and another human?
I don’t mind using an AI image generators for a stupid joke, but I hate seeing them in advertisements. I would never use an AI, local or otherwise, to write something to someone I loved, client, colleague, coworker – but I don’t mind it when I’m trying to wrap my mind around a new API, updating an old program when certain parts I’d written are now deprecated, or even writing something in a brand new language. It’s hard to tell if I’m justifying my uses or if there really is a decent use case for these things.
I guess I’m trying to come to terms with what kinds of AI/heuristic usages I like, am ambivalent towards, and those I hate – to better understand why I feel/think these ways.
- Hi Pete! [↩]
- More than twelve years ago now! [↩]
- TLDR: Take a picture, use Wikimedia Commons, screenshot from video game, lego diorama, use art supplies/image program to make a picture, stickers, pictures of a shelf. [↩]
- I read once that Meta had figured how to associate photos with people based upon the scratches and dirt on the lenses from photos that had been uploaded! [↩]
- This is why I generally rescale/blur/distort images before posting them online. But, I mean, how much time am I going to devote to this if I just want to share a quick picture of something? [↩]