Monthly Archives: September 2022

Little me was never this cute

Posted on by
Reply

Remember MidJourney, the artificial intelligence that turns text prompts into images? Turns out it can also turn images into… more images! So I gave MidJourney a picture of myself from my journal and let it use its imagination. That was… interesting.

Cute little redhead

Pretty sure I never was quite this cute! Although I am sure my mother would not have minded, God rest her soul. She told me a couple of times in my early youth that she had hoped for a girl this time (after three boys) and someone had even congratulated her on finally getting a girl, but that turned out to not be the case. Instead she got me. I didn’t mind hearing that, for by then I already knew that she would have gone barefoot through Hell and back for me if necessary. Not because I was cute, but because she was my mother.

I never had any kids myself. Not only because you still need to have icky, unhygienic sex to make babies (we have the technology to skip that, but most women still insist on doing it that way) but then there would be the daily struggle for two decades to not murder the little monsters, if they were anything like me. Maybe if I had cute kids like this, I would have managed. But let’s face it, there’s no way my little kids could be this cute. And neither could I.

(Machine) learning is not theft

Posted on by
Reply

Hermione Granger by Edvard Munch

Hermione Granger (from the Harry Potter series) painted by Edvard Munch. If you think MidJourney here is plagiarizing Munch’s original, I have a very expensive bridge to sell you.

I have recently mentioned using artificial intelligence to create visual art. Text-to-image applications like DALL-E 2, MidJourney, and Stable Diffusion all use machine learning based on enormous numbers of pictures scraped from the Internet. Now some contemporary artists have discovered that some of their work is used in the underlying database used for training AI, and are upset that they have not been asked and not been compensated.

This reaction is caused by their ignorance, of course. I can’t blame them: Modern society is very complex, and human brains are limited. Yes, even mine. I could not fix a car engine if my life depended on it, for instance. I have only vague ideas of what it would take to limit toxic algae bloom. And to be honest, I could not make my own AI even if I had the money. I just happen to have a very loose idea of how they work because it interests me, because I don’t have a family to worry about, and because I don’t have a job that requires me to spend my free time thinking about it.

Anyway, I shall take it upon myself to explain why you should politely ignore the cries of the artists who feel deprived of money and acknowledgment by AI text-to-image technology.

The fundamental understanding is that learning is not theft. I hope we can agree on this. Obviously, there are exceptions to this, such as industry secrets like the recipes for Coca-Cola or the source code for Microsoft Windows. If someone learns those and uses them to create a competing product, it is considered theft of intellectual property. But if an art student studies your painting along with thousands of other paintings, and then goes on to paint their own paintings, that is not theft. If they make a painting that is a copy of yours, then yes, that is plagiarism and this infringes on your copyright. But simply learning from it along with many, many others? That is fair use, very much so. If you don’t want people to learn from you, then you need to keep your art to yourself. You can’t decide who gets to look at your art unless you keep it private.

The excitement is probably based on not knowing how the “diffusion” model of AI works. So let me see if I can popularize that. Given our everyday use of computers, it is easy to think that the AI keeps a copy of your painting in its data storage and can recall this at some later time. After all, that is what Microsoft Office does with letters, right? But machine learning is a fundamentally different process. The AI has no copy of your artwork stored in its memory, just a general idea of your style and of particular topics. This stems from how “diffusion” works.

When a program like MidJourne or Stable Diffusion gets a text prompt, it starts from a “diffuse” canvas covered in a single shade of color (or grayscale, if a black & white image is requested). It then goes through many steps of moving these pixels into shapes that fit the description it has been given. (It can do this because it has gone through the opposite process millions of times, gradually blurring the images away. Thus the name “diffusion”.) You can, if you have the patience, watch the images gradually become less and less diffuse, slowly starting to resemble the topic of the prompt. In other words, it starts with a completely diffuse image that becomes clearer and clearer. You can upscale such an image and the AI will add details that seem appropriate for the context. (Especially until recently. this could include adding extra fingers or even eyes, but the latest editions are getting better at this.)

It is worth noticing that there is also an excessively long random seed included in the process, meaning that you could give the AI the same prompt thousands and thousands of times and get different versions of the image every time. Sometimes the images will be similar, sometimes strikingly different, depending on how detailed your request is. Once an image catches your eye, you can make variants of it, and these too have a virtually unlimited number of variations.

At no point in this process does the AI bring up the original image, because there are no original images stored in its memory, just a general, diffuse idea of what the topic should look like. And in the same way, it only has a general, diffuse idea of what a particular artist’s style is. My “Munch” paintings certainly look more like Munch than Monet, but it is still unlikely that Edvard Munch would actually have painted the exact same picture. In this case, of course, it is literally impossible, and that is exactly the scenario where we want to use engines like these. “What if Picasso had painted the Sixtine Chapel? What if Michelangelo and van Gogh had cooperated on painting a portrait of Bill Gates?” The AI is simply not optimized for rote plagiarism, but for approximation. It is like a human who spent 30 years in art school practicing a little of this and a little of that, becoming a pretty good jack of all trades but a master of none. They can’t exactly recall any of the tens of thousands of pictures they have been practicing on, but they’ve got the gist of it.

***

As for today’s picture, it was made by MidJourney using the simple prompt “Hermione Granger, painted by Edvard Munch –ar 2:3” where ar stands for aspect ratio, the width compared to the height. This generated four widely different pictures, and I chose one of them and asked for variations of that. This retains the essential elements of the picture but allows for minor variations as you see above. So it is not because the AI had an original picture to plagiarize – I asked it to make variations on its own picture. With some AI engines, you can in fact upload an existing picture and modify it, but this is entirely your choice, just like if you modify a picture in Gimp or Photoshop. The usual legal limitations apply, you can not hide behind “an AI did it!”. So far, AIs are not considered persons. Maybe one day?

 

Suddenly Sudowrite

Posted on by
Reply

Children playing ball, impresionist image

Children playing Calvinball, as imagined by the AI art program MidJourney. Clearly today’s rule is “Bring your own ball”. Luckily today’s main character has that and a spare.

Returning readers will probably not be surprised to learn that I have written millions of words in my lifetime. That doesn’t really take much. I usually write a couple of thousand words at least on an average day when I am not sick, and that’s not counting anything I might write for my job. As you may guess, “writer’s block” is not really my thing, because my writing is like the old house by the river where I lived in 2010, which had three outer doors plus a shed roof you could climb out on from the upper floor. If one of the exits were to be blocked by the copious snowdrifts we had in winter, I could simply use one of the others. And so it is also with my writing.

I am very nearly the worst imaginable candidate, I guess, for the Artificial Intelligence-driven creative writing tool called Sudowrite. It is specifically designed to combat this mysterious phenomenon, “Writer’s Block”, that many writers claim to have experienced. Naturally, I had to try it. (Not writer’s block, but Sudowrite.)

I had read a few reviews (and watched a couple more on YouTube) and they mentioned that you have to apply for access, then after a couple of days you will get an invitation and then you can join. So I signed up, planning to use those couple of days to read more practical reviews and how-tos so I would be prepared when the invitation came. Instead, after I had signed in with Google (Facebook is also accepted) I suddenly found myself on a website that was, in fact, Sudowrite. It gave a very quick tour of the most central couple of features, then left me to my own devices. Luckily there was a link to a (still very brief) Sudowrite guide. But otherwise, I felt much like Galadriel in Amazon’s hilarious new Lord of the Rings parody, where she has rashly jumped into the ocean en route between Middle-Earth and the Undying Lands. Now what?

***

The obvious choice, I thought, would be to copy the not quite 1000 words long prelude to my latest fiction story. In this scene, the narrator picks up a very unnatural-looking crystal that he found embedded in a stone, and immediately falls into the Nexus of Worlds, which is (very obviously, I thought) the user interface to an alien virtual reality simulator that uses Artificial Intelligence indistinguishable from magic to produce a world based on the user’s memories. In this case, he is sent back to 1999, but a 1999 with magic.

Now I tasked Sudowrite with writing a continuation. It proposed two very different passages. I took the first one, deleted the crazy plot twist, and edited the rest. Then I wrote my own short continuation, introducing the Ultimate Book of Magic which is the central item in my actual Work in Progress. I asked my new friend Sudowrite to describe the look of the book, and I actually kept most of that. Sudowrite is really good at feminizing novels by proposing all kinds of sensory information, going into detail on how things look, feel, sound, smell, and taste. In case you wonder how the Ultimate Book of Magic smells, I can tell you now: “The worn leather smelled like a library. Like the smell of wood and old paper. The book smelled of mold and dust. The metal clasp like rusting iron and blood.” And should you be so lucky as to get your hands on it, you would notice that “The cover was worn and smooth. If I had to guess I’d say it was oiled, but there was no sheen to it. It looked like it had been oiled a thousand times.”

(I am told women love that books contain all kinds of sensory detail. I noticed it first in Clan of the Cave Bear, which contained more information about Ice Age vegetation than my encyclopedia at the time. If I were to add stuff like that, my books would be thousands of pages long. Y’all know how verbose I can be even without that kind of peacock tailfeathers. In all fairness, it is not like all male writers excel at self-limitation. There is, one might say, no such thing as a tad Williams.)

Anyway, Sudowrite and I continued taking turns writing a couple of paragraphs each. I would delete wild plot twists, edit the rest, then try to steer things back on track in my own paragraphs. It quite feels like trying to write a collab with Calvin from Calvin & Hobbes. There is always a lurking sense of Calvinball, defined from the horse’s mouth: “Other kids’ games are such a bore!
They gotta have rules and they gotta keep score!
Calvinball is better by far!
It’s never the same! It’s always bizarre!”

You may as well memorize this little verse before you start writing fiction with Sudowrite. Or nonfiction, for that matter, because Sudowrite will play Calvinball there too, unlike the various AI writers that are tailormade for writing ads and paid blog posts. (Rest assured this post is NOT paid by Sudowrite or even their competitors.) Quote Techcrunch: “Asking Sudowrite to describe what a startup is had me laughing so hard I was gasping for air.” Yeah, I can imagine. That is, after all, the purpose of Calvinball. And Sudowrite is nothing if not Calvinball.

That said, it is a true Artificial Intelligence. The more you work with it, the more it gets to know you (and the other way around). Take the following sequence, where you can hardly see where one of us leaves off and the other takes over:

“Apart from the proper Sigil, and the correct postures and incantation, the Affinity of the Binding is limited by the quality of the Exemplar – the object symbolizing the Source – and the mage’s natural Resonance with the Source.”
“If the mage’s Resonance is weak, the mage will need to use an extremely potent and pure Exemplar. If the mage has a strong Resonance, they will have more leeway in choosing an Exemplar, and if the mage already has a strong Affinity with the Source, simply having a properly prepared Exemplar may be enough.”
Hooray for hyperlexia!

Here the first paragraph is by me, while the explanation is by Sudowrite, except for a single word added. And yes, it was Sudowrite that wrote “Hooray for hyperlexia!”
Shut up and take my money, Sudowrite.

***

I may not actually use this in my writing (except perhaps during NaNoWriMo) but, in the winged words of Sims 3: “Magnus is having so much fun it is almost criminal”. Sudowrite is not going to write your next novel for you, but it can help you create new ideas, new characters, new plot twists, and descriptions varying from the mundane to the ridiculously elaborate, depending on what tone and style you prefer. Personally, I am keeping my Sudowrite experiment separate from my current writing project, but ideas are good climbers.

For those who have been working on their Great American Novel for twenty years and take it very, very seriously: This is not for you. Madness is not the only danger in writing: There is also the danger that something may be understood that you didn’t want to know. Like, that writing can be fun. But as for me, I already knew that. I am never lonely when I have my invisible friends in my head. Sudowrite is just another disembodied companion joining my brainstorming sessions. (But possibly the most hilarious one.)

Will I become an Artificial Intelligence?

Posted on by
Reply

This painting of a west Norwegian farming village is actually made by an artificial intelligence named MidJourney. And indeed, it is a sight many of us west Norwegians would not bet against having seen in our travels. It is not copied off the Internet though: I watched the AI go through several versions of this painting before I settled on this one.

Today, September 10th, 2022, was the first time I heard a YouTuber greet his audience as “human and AI”. I wonder if that will be the new “ladies and gentlemen”? Maybe the time is finally drawing close when, as a much better man than me once said, “God can wake up children for Abraham from these stones.” I have already long had in mind when writing my journal, that someday after my passing, an AI may read through Google’s archives and find my excessively detailed thoughts there. But maybe it won’t even be after my passing, if I take my vitamins and cut down on the Pepsi. You see, the progress in the field of AI has been nothing short of remarkable recently.

I returned to this topic of interest a couple of weeks ago after a casual mention on a website of a new program that generates images from text prompts. Turns out there has been released a flurry of such image generators this year, and they are growing steadily more advanced. My current favorite is MidJourney, and it is probably the most popular these days. After the latest upgrades, it gets faces right much of the time, at least when it gets to focus on them. This has proven to be one of the most difficult parts for AI, with the notable exception of the AI that specializes in them. Yes, “This person does not exist” which provided my picture in October last year, is also an AI.

So this is not my first run-in with Artificial Intelligence, far from it. Grammarly, a spelling, grammar, and style checker that I use under doubt after it has been greatly improved, is also an AI. It still makes some embarrassing mistakes, but then again so do I. Together we do better.

In my archives, you will find a number of versions of Dragon NaturallySpeaking, the speech-to-text program that eventually became better than nothing. I used it for years when my wrists hurt worse than my throat, but these days it is the other way around so I rarely fire it up. It is pretty amazing now though, and probably even better if you have English as your first language. Actually, that might depend on your dialect, I guess. According to Dragon, I speak Great Lakes English. I am pretty sure they don’t sound like the bandits from Skyrim though, while I do. Anyway, that’s another Artificial Intelligence application. In fact, it was based on the work of Ray Kurzweil, the great prophet of AI and modern futurism in general. And he is indeed planning to become an AI of sorts, by uploading his mind to a computer when the technology is ready for that. Good luck with that.

***

Anyway, today I was planning to demonstrate for y’all an AI tool that can take a brief outline or just a bunch of idea keywords, and expand them into a full-sized blog post. Doesn’t that sound amazing? But given that all the paragraphs above were my introduction before I started on the main topic… maybe I don’t need it. On the other hand, maybe we all (if we live long enough) will begin to use AI in so many facets of our life that we gradually become them, without ever noticing.