• Hi Guest!

    The costs of running this forum are covered by Sea Lion Press. If you'd like to help support the company and the forum, visit patreon.com/sealionpress

Makemakean Does Various Graphical Things!

On this point, one of my many "things that I want to do some days" would be some sort of Wikipedia-like website (though more curated, like nLab) that would serve as a home for illustrations of abstract concepts in mathematics. The idea would be that people who are researchers and scientists who have an idea for an illustration for something, but aren't necessarily the best of artists could draw a basic sketch of their ideas to the best of their abilities, and then people with more artistic skills could take over and develop it further, and there'd be continuous discussion and feedback, the point being that over time, we'd get a lot of good, free-for-all-to-use illustrations of these various concepts.

I spoke with one of the guys running nLab some time ago about this, and while he says he thinks it's a great idea, he did advice me that there would be a lot of infrastructure involved in running something like this.
 
I don't doubt I would from the few scans I could find online. Mainly though, I had Sven Nordqvist in mind when drawing this (who I mentioned in my Top 11 illustrators), though he is of course far more detail-oriented than me. He has drawn these books which essentially just follows one or two characters as they go through a variety of landscapes and rooms, with each drawing leading directly into the next:

Hundpromenaden%2B2.jpg
Yeah, that's the same way It's For You is done, starting with the protagonist opening a box to find a whole world inside, and it ends many connected 'rooms' later with him finding the original box at the end - hence why it came to mind.
 
On this point, one of my many "things that I want to do some days" would be some sort of Wikipedia-like website (though more curated, like nLab) that would serve as a home for illustrations of abstract concepts in mathematics. The idea would be that people who are researchers and scientists who have an idea for an illustration for something, but aren't necessarily the best of artists could draw a basic sketch of their ideas to the best of their abilities, and then people with more artistic skills could take over and develop it further, and there'd be continuous discussion and feedback, the point being that over time, we'd get a lot of good, free-for-all-to-use illustrations of these various concepts.

I spoke with one of the guys running nLab some time ago about this, and while he says he thinks it's a great idea, he did advice me that there would be a lot of infrastructure involved in running something like this.
That would indeed be a good idea, it's tiresome at best to either try to make your own when it's not your forté or to hope something you found on teh interwebz won't get you sued.
 
That would indeed be a good idea, it's tiresome at best to either try to make your own when it's not your forté or to hope something you found on teh interwebz won't get you sued.

That is another good point that I didn't really think about.

I do think that seeing that there are talented people out there who are willing to spend hours and hours for free making election maps for Wikipedia, it would strike me as odd if there weren't talented people out there who would be willing to spend hours and hours making scientific schematic diagrams for free, in particular if one could decide upon a standard style, cetain guidelines, and the stuff.

Problem is of course that the reason why people are willing to do this for free is because they know people will see their work. That is, once one has the website up and running, then it would be self-sustainable (in some sense), the problem is getting it there.

The second problem is figuring out how to properly set it up in the first place, what the format should be, since the focus, the point of it all, is supposed to be about illustrations, not any accompanying articles explaining the whole thing.

In a sense, I would want to have it be both a version of StackExchange and a version of Wikipedia, which begs the question of, "Well, what does that mean concretely speaking?"

This would of course be something I'd first start looking into once I have my PhD, and so not something in the immediate future. I'll have to think about it some more.
 
That is another good point that I didn't really think about.

I do think that seeing that there are talented people out there who are willing to spend hours and hours for free making election maps for Wikipedia, it would strike me as odd if there weren't talented people out there who would be willing to spend hours and hours making scientific schematic diagrams for free, in particular if one could decide upon a standard style, cetain guidelines, and the stuff.

Problem is of course that the reason why people are willing to do this for free is because they know people will see their work. That is, once one has the website up and running, then it would be self-sustainable (in some sense), the problem is getting it there.

The second problem is figuring out how to properly set it up in the first place, what the format should be, since the focus, the point of it all, is supposed to be about illustrations, not any accompanying articles explaining the whole thing.

In a sense, I would want to have it be both a version of StackExchange and a version of Wikipedia, which begs the question of, "Well, what does that mean concretely speaking?"

This would of course be something I'd first start looking into once I have my PhD, and so not something in the immediate future. I'll have to think about it some more.
Case in point, I've just spent a few hours updating slides and making a lecture handout, and hoping "from Wikimedia Commons" is sufficient attribution for some diagrams.
 
So I finally got my invite for Dall-E, and, well, it truly is an insanely powerful piece of software. First thing you can do is to type in prompts and see what it produces. Seeing I've happily been following @Ciclavex 's All The Presidents Shuffled, of course I had to type in "Ulysses Grant as Founding Father", which truly rendered some badass-looking imagery, albeit not necessarily particularly area-accurate:

DALL·E 2022-09-29 01.26.53 - Ulysses Grant as a Founding Father.pngDALL·E 2022-09-29 01.27.01 - Ulysses Grant as a Founding Father.png

And somewhat more, err-... gorilla-ey looking imagery:

DALL·E 2022-09-29 01.26.40.png

Okay, so it sometimes misfires, but it still looks deeply professional when it misfires. Sometimes it backfires in truly bizarre ways, like when I typed in "Oil painting of 19th century emperor of Scandinavia" and it gave me this:

DALL·E 2022-09-29 01.27.23 - Oil painting of the 19th century emperor of scandinavia.png

I am amazed that we have gone so far in our development of AI that it's now gone and become woke, both colour-blind and gender-blind in its casting of characters! That's certainly something! :p

The AI could also generate variations of pictures you plugged into it. So I plugged in photos of myself. The result was... frankly incredibly creepy:

DALL·E 2022-09-29 01.25.45.pngDALL·E 2022-09-29 01.25.38.png

I think I'm going to have nightmares about those two guys stalking me.

It really misfired when I asked it to try to make variations of hand-drawn drawings of mine. Take for instance this picture of Princess Ernelinde. Delivers this:

DALL·E 2022-09-29 01.26.26.png

Horrifying.

But, what truly impressed me was how good it was when I asked it to produce variations of my vector art image of Rome that I made this summer. It provided me with this:

DALL·E 2022-09-29 01.28.23.png

...and...

DALL·E 2022-09-29 01.28.52.png

...and I must say, those are some beautiful things.
 
I’ve been trying DALL-E for a month myself. It rarely generates what I want, but it’s always interesting. And only sometimes viscerally horrifying.
 
I cannot stress how insanely good this AI is! So much so that I'm worried about the future of humanity! Whereas Dall-E took this:

index.php

...and delivered this:

index.php

...now look at what we're getting:

the-most-beautiful-princess-in-the-whole-world-trending-on-artstation-sharp-focus-studio-photo...png
 
I also fed it this picture of Emmy:

index.php

and it delivered this:


...zoom in on the face and ask it to increase the resolution:


Like, wow! Wow, truly, wow!
Not to harp on about it, but that one literally does look like it comes from this gallery.

TBF Marvel just got away with a Thor film in which the villain is literally copy-pasted from the same book series, so you're probably fine.
 
Not to harp on about it, but that one literally does look like it comes from this gallery.

TBF Marvel just got away with a Thor film in which the villain is literally copy-pasted from the same book series, so you're probably fine.

Feel free to harp on, though I must admit that I don't think I'll read Sanderson any time soon, seeing I don't want to be too influenced by his conception of Shallan. I have my own idea of what kind of a girl Emmy Nicander is, and I kind of want her to be able to grow organically on her own. It's not the legal implications that bother me (if indeed I ever get far enough with this to ever actually publish anything), it's the whole thing about being able to feel that she is sort of "my own character" that is very precious to me.

I on the other hand will harp on relentlessly on how insanely powerful that AI is.

You can put in your own stuff, have it put together something with a given degree of image retention, then you can take what you got out, modify that in Photoshop to make it fit your desires, plug that thing in again, and on and on and on...!

Getting ever closer to what precisely it is that you desire!

Like, I'm pretty pleased with my Photoshop skills, but they have never been for the sort of quasi-photo-realistic stuff you see people put up on DeviantArt. I look at that stuff and I marvel and envy the creators behind them. I can do cartoony stuff, and I can do cartoony stuff reasonably well, but that's about it.

But thanks to this AI, I have been able to produce this depiction of Nat Van Oosterzee/the Isaurian such as I imagine her looking like around 1876:

Isaurian-ca1876-smaller.png

I do love how because of how the AI works, she looks slightly like so many, many people, and still no one of them in particular. I was also able to tweak and experiment enough to get a picture of what she looks like come the 1920s:

Final_Isaurian-ca1920-smaller.png

I would assume that as the late 19th century was Steampunk, so the 1920s must be, well, the Roaring Twenties on steroids and with computers? I dunno, I'll think of it eventually.

EDIT: Also with as she appears c:a 1964 and as she appears c:a 1997.

An Isaurian for all Ages.png
 
Last edited:
EDIT: Also with as she appears c:a 1964 and as she appears c:a 1997.

This AI ain't entirely perfect. They don't all look exactly like the same woman. And on occasion, you can sort of see certain individuals whose faces the AI is drawing from as it is trying to execute my requests. But that being said, it is still an insanely, insanely, insanely powerful AI, and if you take what you get out from it, modify the result in Photoshop, plug it back into the AI, and keep doing so and again and again and again, this baby can deliver some insanely good results. So much so that I would kill to get to have a look at the source code behind it. Not because I want to steal it as much as because I want to admire the pure ingenuity that is behind it.

Oh, well.

COVzyag.png

No doubt there's some very interesting story behind each and every one of these pictures, and one day, I'll figure out what they are.
 
Last edited:
On the topic of AI's and how good they are at drawing stuff, I'm still convinced that for many years to come, there will still be a need for artists, and my reason for thinking that is based entirely on a short story in Svålhålet by Mikael Niemi, in which they hook up the great intergalactic internet across the entire universe, and together with universal translators, they soon enough discover that everything possible to write has already been written somewhere by someone.

This causes problems for the authors on Earth at first who become extremely despondent since they can no longer get any copyrights, but it turns out that in the end, they are still being requested. People still want them, and they soon become celebrated in their new jobs. They simply become librarians, because they know how to find the good novels out there.
 
About the AI, I have since found out that the AI in question--Stable Diffusion--is in fact Open Source software, which explains why the people behind the website can offer 1,000 free generations a day, whereas DALL-E can only offer 20 free generations a month. This means that I actually can look at the code if I want to, and I much look forward to doing that in detail some day, as well as reading the documentation. (Though in all probability, it will soon prove to be quite, quite beyond me.)

Anyway, to generate some non-Isaurian pictures. A few years ago, I made this watercolour of Police Master Strömer from The Great Nordic Election Night that I was very pleased with:

Strömer.png

I plugged that into the AI, and got out this:

Stromer1.png19th-century-policeman-haggard-trending-on-artstation-sharp-focus-studio-photo-intricate-detai...png
 
Back
Top