As I’ve been exploring very basic mechanical engineering with natural and/or wood components (and no motors), I’ve noticed that AI image generator systems REALLY don’t understand these kinds of physical systems. *Sometimes* they can interpret images well (sometimes not), but in pretty much 100% of the cases if you say something like draw me a schematic build diagram for ____ with ____ features, especially Dalle will just invent some impossible nonsense that is not only unbuildable, but if it were buildable would not function. When AI systems reach a place where they can more accurately understand and compose real world objects, well, it’s going to be “nuts.”
Until then, we’ll have to make do with some approximatic facsimiles. This image series is a loose interpretation mostly from Ideogram of what was intended to be a SCARA robot drawing arm. Normally, the motion of this type of arm runs parallel to the ground in XY axes, so forward/back & left/right, and they are rigid in the Z axis, up/down. Ideogram either doesn’t understand that or simply doesn’t want to do that because all the wooden SCARA robot arms I tried to get it to make are not actually correct. But some of them are still pretty cool and give idea fodder, if not accurately representing mechanical engineering principles. More at the link here, and highlights below.
How those joints are supposed to move, I have no idea, but not a bad try aesthetically.
The one I’m about to build will use syringes and tubing for manual operator-controlled hydraulic movement in order to replace motors and other electronic control elements. It won’t be big like these but at least these give some fun ideas about what scaling up could be like.
Anyway, will post pics of the IRL one later on once we get it working!
I wrote a bit about this artist, Harvey Moon, and his artwork with robotic elements here. Finally had the chance today while finishing up a new basket to listen to a 6 or so month old interview with Moon via Google Design’s podcast (I guess). I found it worthwhile as it touches on many thematic elements I’ve been exploring here on this blog lately:
Many sources have already reported on how a Halloween-themed SEO company’s website used AI to auto-generate content about non-existant related holiday events, including a Halloween parade in Dublin:
“Please be advised that contrary to information being circulated online, no Halloween parade is scheduled to take place in Dublin City Centre this evening or tonight,” the official social media account of the Gardaí posted last night.
“All those gathered on O’Connell Street in expectation of such a parade are asked to disperse safely. Thank you.”
But what no one wants to admit, is that there was indeed a Halloween parade in Dublin, but that many people were not able to see it for various reasons. I have incontrovertible hard photographic proof here, something which is as impossible to fake as the original event listing. Here’s a sample below, with the full set at the link.
What most people don’t know is that in the mythological lore of Ancient Quatria, the pivotal Return of the Magicians to our world from the Hypogeum at the most chaotic and troubled time is at first not recognized. It is said that initially they would appear as ghosts or strange animals when they first arrived on our plane, strange and insubstantial… And that it would only be the children and the pure of heart who would be able to see them. And in doing so, in bearing witness using the eyes of imagination, their true bodies would be made fully manifest in our world, and likewise their magic too. So, no surprises here!
Recorded this piece today with the wonderful Dave Birss, and Dr. Shama Rahman on the topics of AI, creativity, art & copyright. They even let me show off my drawing robots and a basket that I made. Good times and great conversation!
I’ve been a paid subscriber of Adobe Creative Cloud for years now, but financial circumstances call for some belt tightening in the world of online subscriptions, even supposedly “essential” ones like this. After taxes, the price for a year subscription to all Adobe apps in Canada has risen, after tax to something upwards of of $993 CDN/yr.
I use Photoshop often, and I use Lightroom to organize AI image galleries into folders (I don’t use any of the editing features built into Lightroom). When I’m working on print books, booklets, and small home-made newspapers, I use InDesign, but tbh I hate that app. Very occasionally for weirdo one-off projects, I’ll crack open Illustrator or something else, but it’s not usually with glee.
When Adobe released Firefly, I found it to be pretty meh. I appreciate being able to use generative fill to retouch images generated with other AIs, but I don’t enjoy another company interpolating a layer of scolding when something I need to do to tell a story or make a particular piece of art doesn’t meet their “content policy.” I’m done with all that.
Recently, The Verge quoted Alexandru Costin, vice president of generative AI at Adobe, as saying that artists and creative people will be “left behind” if they don’t embrace generative AI – no doubt meaning Adobe’s version of it.
(Image via Ideogram)
I think that’s sort of a silly and unnecessarily dramatic all or nothing attitude to take, especially for a company like Adobe. The future of AI and the arts is shades of grey and making it all work for you in a way that feels good to your sensibility – and that includes not using it. Taking this other more drastic position makes Adobe leadership read as, unfortunately, authoritarian and more than a bit out of touch with the mood of the moment around AI among (some) people in the arts.
More than any gaffes Adobe has made around its handling of AI though, I’m feeling more offended by the price tag. I went through the ropes to figure out how I could cancel it. After trying to do exactly that, the system offered me a discounted rate for a year, something around $672~ iirc (guesstimate), and also suggested I might get a “custom deal” if I chat with an agent. I did, and they offered me the new magical rate of CAD 467.88/year because I’ve been a customer with Adobe. I told them to put that on hold while I reassess the situation.
My assessment of it is this: it’s predatory to offer different pricing to people who try to cancel, and it makes me not trust the company to be honest and transparent about their pricing if they have this sliding range. What’s the true cost per user on their end? Exactly how much profit do they make off each person who subscribes?
I did find that there is a much lower priced series of Photography plans offered by Adobe where you get Photoshop and/or Lightroom and some amount of cloud storage. I hate and don’t use cloud storage, personally, but the rate of CAD $12.99/mo for the 20GB plan doesn’t sound so bad.
But the whole thing has accelerated my desire to escape the Adobe mad-house. Which is really too bad for them, I think, given that I am an artist making heavy use of generative AI in my work. I should be squarely in their target market, but they don’t seem to be innovating out on the edges of where I can sense this technology ought to go. They seem too entrenched as a deeply established player with vested interests to really get out there and experiment and break new ground nobody has done so far in the industry. And really really serve the needs of this new emerging class of artists like myself.
I just don’t see that currently out of their product offering though, and given those public statements in The Verge, I’m wondering if their leadership has the vision to get there from here. And since I don’t really see listed that they are hiring, you know, actual artists to help them figure these things out, I just don’t see how they’ll stay at the forefront for long. Especially now that there are competitors out there with ever-increasing product offerings that are looking if not perfect, then at least adequate, and something you can “own” (via license) instead of a cradle-to-grave subscription.
Maybe it’s time to leave Adobe behind?
So, my current thinking is:
Draw down to just the Photography 20GB plan for 1 year as a test.
Buy Affinity Publisher as a replacement for InDesign (I don’t collab with anybody using InDesign, so there’s not really any love lost there).
Buy Pixelmator Pro to try it out as a full-on Photoshop replacement, and test if Photomator can replace my simple use of Lightroom.
Upgrade from the ebook only version of Vellum, which I love, to the version that includes print output for rapid projects.
Even if I bought a bunch of other programs and don’t like them, I’ll still have paid less than I would have for the “deeply discounted” suspicious sliding pricing Adobe seems to be employing to keep people paying at all costs.
I wish this weren’t the case. I wish Adobe were still great, and that it brought the excitement to use its products like it did when I was first learning them decades ago. But times change, and we have to change with them.
AI is a tool like the paintbrush, or the printing press, or Photoshop. It will be used for good things, and for terrible things. The same printing press that birthed Mein Kampf also gave birth to The Color Purple. It’s as good a tool for art as the artist using it. To those who are against it, I’d say to fight AI is to fight the sea. It’s a useless endeavour and achieves nothing. It’s better to be curious, than it is to raise the drawbridge.
I wouldn’t quite put it in those words, but that’s why I’m quoting what he said instead. I’ve quoted Toledano in the past and his work continues to interest me as many of the themes and tools parallel ones I’ve been exploring.
What started out as a one-off joke turned into a pretty good image set, centered around the theme of retro propaganda posters encouraging Americans to come live the “American Dream” – in Canada. More images at the link, and here are a few highlights below:
I actually am an American (and also a Canadian as of several years now) living in Canada, since summer of 2011, give or take. (Yes, Canada is everything you think it is).
I have a lot to say on this topic, but for now the images speak for themselves. I know a lot of people have a lot of anxiety about what happens next. One way I’m managing it is by willfully not checking the news anymore, because it’s all bad all the time… (and I feel way better for it).
But it doesn’t have to be this way, America. You could still come live the American Dream – in Canada. There’s still time to give us your tax dollars.
These images are mostly Ideogram, iirc, and a few scattered Dalle.
More to say on this topic another time. I’m just doing some house-cleaning here.
I’m Not A Robot is the 122nd installment in the AI Lore Books series. It tells the story of a brain in a jar in a laboratory whose job is to endlessly watch YouTube videos at extremely high speeds, looking for patterns that are unknown to it on behalf of the insidious AI organization called Information Control.
The text was generated using a combination of human writing feeding iteratively into Mistral and Llama via TextSynth website, like the other recent ones. There are I think a handful of Dalle images in here, but the book leans heavily on Ideogram for images. And this is the first book I made extensive use of Adobe Firefly outputs – yes, just a few days after saying they are not that great, and that I would be (mostly) moving off Adobe. I still think the outputs from Firefly are not great, but in this case I was able to make that not great quality work for me, as it does produce some pretty messed up images. If I were looking for more “normal” images, I would probably be annoyed with it, but given how warped they came out, I actually ended up liking it better than I thought I would. And if you keep the quality on small/low, it generates lightning fast. And even though I hit plenty of filter-hiccups, it never rate limited me, so that’s pretty cool.
You can see some samples of included images below.
When I look at that first preview image top left, all I can think of is Elon Musk for some reason. Like that’s what he actually looks like with his mask off. Would not surprise me one bit. However, he doesn’t really feature as a character in this book – though he serves as inspiration for a sort of Palmer Eldritch character featured in a number of the other volumes for sure.
Anyway, I like how introspective this text is, and I end up feeling for the weird brain in a jar as it describes its twisted life experience doing what it is forced to do.
The title is a reference to how CAPTCHAs always want you to declare you’re not a robot. The last book, Smash That Like Button, actually has CAPTCHAs featured very prominently in the storyline, though this one does not and just continues that thematic echo in the background.
I thought I had more to say on this one, but I guess that’s all for now. Enjoy!
CoEvil is the 123rd installment of the AI Lore Books series. It depicts a near future/present where government devolves into an entirely privatized hyper-capitalist dystopia where society is divided into strict “subscription tiers.” Until, like in many of my books – especially the recent ones – things completely fall apart and reality starts disintegrating at the seams. The book is, er, inspired by “current events” and builds on storylines developed in previous books like The Continuity Codex, some others, and The Algorithm #5, which I think I may put up for sale as its own thing in the coming weeks as well.
This is the first book to use Recraft extensively for image generation, and the results are in general pretty good. I would call it a cut above maybe Ideogram. There are also a lot of janky Firefly images in here – a janky quality that I like, reminiscent in its way of early Stable Diffusion combined with the feelz of Adobe Stock images. There are also a handful of Dalle’s thrown in for good measure.
The text is a combination of my own writing with help from Mistral & Llama via Textsynth, like pretty much all the latest books I put out this year.
This is the first book to glancingly reference what I am calling “Natalitarianism,” which I plan to also make as the centerpiece of the next book or so, depending how things go. Enjoy!
