WHO IS AN AI-ARTIST? AND HOW CAN YOU USE ARTIFICIAL INTELLIGENCE TO FULFILL YOUR DREAM.
For the cover for this article, I asked @DEADWEIGHT to send me his picture. I first drew my graphic on it, and then combined it with the original photo using AI. Then I tweaked it a bit more.
Hey, guys, today I want to share with you a very interesting conversation with an AI artist. Recently I came to "CryptoArt Ukraine" telegram-chat room and was very positively greeted by the participants. We had a cool and interesting conversation and in the course of it I got a couple of questions to one of the participants - @DEADWEIGHT. We met in a private chat and now you can watch our discussion below. Enjoy. (...and please remember that the author of the article is not a native English speaker, there will most likely be mistakes in the text, if you see them, feel free to contact me personally on Twitter - @igrushkina1, I will correct everything).
CryptoArt Ukraine chat : https://t.me/joinchat/WI6AFR_KiGgsCMCd
You can find all of @DEADWEIGHT's accounts at this link and learn more. https://linktr.ee/weightdead
"Lobsterhead" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/281476
IGRUSHKINA: Before you started doing AI art, did you have anything to do with art in general? I wonder how you came to work specifically with AI and what you were doing before that.
DEADWEIGHT: Before that there was music, poetry, photography, video and now AI art.
IGRUSHKINA: It's quite extensive. A wealth of experience. Have you been doing all this professionally? Or is it more of a hobby?
DEADWEIGHT: As a hobby it's always been, it's just that in different periods I've been more involved in a particular kind of creativity. There were a few years when I wrote poetry. But for the longest time I was involved in music. First there was a music band, then I started writing electronic music on my own. I took photography for a very long time. I started with digital photography, then came to film, and then went back to digital again. There were a couple of projects where I shot 640x480 pics on a 0.4 mpx button phone.
I was tweeting everyday snapshot and printed these tiny pictures and gave them to my friends and strangers.
For example, you can check it here:
https://twitter.com/i/events/847784021572677633
https://twitter.com/i/events/881515597800013824
https://twitter.com/i/events/811202819407626241
https://twitter.com/i/events/907620428755193859
One of a series of daily photo cards by @DEADWEIGHT
IGRUSHKINA: Was there some kind of conceptual idea behind it? Or did you just decide to take pictures every day and hand them out to people? It seems to me that the artist is always searching for a way to express himself, to convey experience, his ideas... Maybe you wanted to do something to express your interest in a topic and that was your search for self-expression?
DEADWEIGHT: Yes and no. For example, in the case of AI, it was primarily an interest in technology. To what can be done here in different applications. The aesthetic effects of what comes out. I wanted to play with the technology, get to know it. And then to already produce meanings and contexts. Here, the most interesting and mysterious thing is the synergy with the machine, rather than what you can do yourself. This is the interest of this technology. It produces new meanings, readings, aesthetic patterns and visual effects with its results. So the expression of meanings was important, but rather secondary. Primary is the researcher's interest in a new instrument for self-expression. Like a brand new musical instrument, which has just appeared on the market, but no one knows how to play it yet. It's very exciting.
And expressing meanings is to the point about how I dreamed of drawing but couldn't. Now, with a machine I can draw, but it's not at all the same drawing as academic artists do. By the way, thanks to all the generations of artists that neural networks are trained on. Without them, the machine wouldn't know how to draw.
"Jesus Christ Superstar" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/344319
IGRUSHKINA: Academic artists are one direction. There are many artists in the contemporary art field who have nothing to do with academic painting, they are interested in other aspects. There's an interesting question here. Even if the artist is not a follower of the academic school, it is still own self, not his instrument that generates the message of his work - it is his idea, his skill. In the case of AI, it turns out that it is the tool that learns to visualize on a given theme. And the theme is set by the one who "runs the ball". So it turns out that the role of the one who "runs the ball" is the role of the teacher?
DEADWEIGHT: When working with AI, in many ways, I am guided by the already formed experience of interaction with the neural network. So in our tandem, I am the one learning, not the neural network. As far as I know, it is pre-trained and does not trained in a particular case. I always considered myself an artist, and I still do. Because an artist is not specifically about drawing, it is about producing aesthetically and semantically valuable works/artifacts. The role of the teacher is performed by people who train neural networks, feeding them various databases and configuring their work. That is, they are mostly programmers. But if the artist oneself does it, then in that sense, yes, he becomes a teacher. I have not yet been involved in feeding datasets to neural networks, but I really want to do it when I'm ready for it.
IGRUSHKINA: From our conversation, we can understand that although you have been associated with art, you have never drawn before? How did you come to the point where you want to make art with AI in the first place?
DEADWEIGHT: No, I didn't draw, but I've always wanted to. I've always been interested in cutting edge technology and futurism. I like to observe technological progress. I find a lot of positivity and beauty in it. I started following the development of AI about 5 years ago, it looked very interesting and promising. AI technology has evolved a lot in that time. I looked at all these advances and envied people who could deal with AI. For me, the barrier to entry was not knowing how to code. But I was itching to start interacting with AI myself. The potential for co-creation with AI is amazing.
"Astronauts" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/402813
IGRUSHKINA: So AI is a kind of salvation for you. Considering that you've never drawn, but you really wanted to, and now you can create art with AI. You control it, you guide it, and it "draws" for you.
By the way, can you talk in detail about the process you're involved in as an author interacting with AI? Are there any difficulties in this process, any problems you're solving. Does something work, something doesn't, any experiments maybe?
DEADWEIGHT: I've been lucky. Already with the second generated picture, it was clear to me that this is just WOW. In a way, working with text-to-image is a lot like a magician pulling a hare out of a hat. You never know for sure what you're going to get, even if you work with it pretty closely. I like chaos and randomness as tools for creativity. There's a lot of that in neural networks. At first it was hard, because you get only 5 credits right away (we're talking about NightCafe studio), you can use them for 5 generation attempts. Each day you get 2 free credits. The rest - for money. And to work with this technology, you need hundreds if not thousands of attempts to weed out the unnecessary and leave only those works that are of artistic value.
If we're talking about my artistic experience, here, for example, is my second picture that came out together in work in AI: https://creator.nightcafe.studio/creation/vqyHWAQ4bY6CEzdoqV3H
"Unfollow the White Rabbit, Neo!" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/344235
IGRUSHKINA: Cool rabbit). As I understand it, the whole process was about you writing the phrase and getting the result suggested by the neuron. You basically only influenced it by asking the machine - "So, generate me something randomly about the Wonderland rabbit and there should be a Neo from The Matrix here as well".
DEADWEIGHT: Yeah, that's about it. The main magic happens in the selection of the query. In this work, I was wondering how the machine would portray the query "Unfollow the White Rabbit, Neo!" Especially in the "Unfollow" moment. I didn't know how AI understood syntax and semantic blocks.
IGRUSHKINA: And in your opinion, how did the neuronics handle this task? What did the result tell you. Do you somehow use the lessons learned from each result in the future?
DEADWEIGHT: More than that. I didn't know what I was going to get at all and so I was surprised by this kind of "grumpy rabbit" with matrix elements in the background. As far as I see, the neural network VQGAN+CLIP, which is used in NightCafe, is quite bad at rendering faces, and here we got such a wild thing! "That hare doesn't bode well, unsubscribe from it urgently, Neo!"
I take a lot of pictures and pay attention to interesting effects. I analyze. I get some interesting results because of this, and I try to combine them with others. I already have some favorite techniques. On the one hand this is great, but on the other hand it hinders my development, when I can walk in the paths I have explored. Therefore, sometimes I purposely try new methods in order to achieve new and interesting results.
"You've Been Downloaded" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/482918
IGRUSHKINA: You say that you'd be interested in training the neurons themselves, teaching them to perceive your requests, not just learning how they respond to yours. What's stopping you from diving into that?
DEADWEIGHT: This question gets bogged down in different factors. From the inability to code and lack of time for all the interesting activities, to the weak hardware on my computer. So, in theory I have this idea, but it is far from being at the top of my priorities. Some things I don't rush, they have to mature like fruit on trees. I've been waiting for an opportunity to start working with neural networks for about five years. It happened this summer, when I came across a link to the site NightCafeCreator. It's the easiest and most pleasant for me way to get into the technology.
IGRUSHKINA: Okay, we've come to what you've always wanted to:
b) work with neural networks...
This summer you found out about NightCafeCreator. And now - it is your tool, with which you draw and work with neurons. Now we can talk about the process of working with AI.
We've already talked about this topic in the "CryptoArt Ukraine" telegram chat with you. You said that it's a creative process. It's like magic - a pinch of this, a pinch of that... and with each time you start to understand how the machine "thinks" and it starts to understand what you want. You try different options and look at the results. Those results are kept in your profile, and later you continue working with what you have got, trying things out, and adding more. The machine remembers and reacts. For example, it understands what you liked best, and then you continue working in that direction, experimenting. Or is every time a completely random result?
DEADWEIGHT: In my experience, if you enter the same text query (prompt) and generate new and new work, the results will generally be similar. They will differ, but not very much.
IGRUSHKINA: Interesting. I thought AI is learning all the time. Each time it learns what's good, what's bad, what requests users make, and whether they're happy with the results. Maybe it even remembers how people react when they repost an artwork, what gets more likes, what comments they write.
DEADWEIGHT: That would be nice, but as far as I know, it's not. I think it's a matter of the near future. Maybe, of course, it's already there, I just don't know it. Things are evolving very quickly now.
"Found in the Santa Muerte Carnival" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/542036
IGRUSHKINA: I began to have this understanding. Working with AI as working with art tools is the same as when an artist works with oil, for example. He learns to feel the structure of paint, how it lays on the canvas, to try to mix colors. In general, artist learns how to use an artistic medium.
Do you only use NightCafe? There are now many different services for working with AI, aren't there? What do you know about it?
DEADWEIGHT: It's kind of random, it's just coincidental. Before I started using NightCafe, ArtBreeder already existed. But I never saw it or I just passed it by. I tried it after NightCafe. Very interesting tool, I like in it much more tangible and clear system of generation control, which is more like a plugin for Photoshop than the Google query bar, where you type in "Unfollow the White Rabbit, Neo!" and pull out a bunny. ArtBreeder has much more control over the generation process. In that sense, text-to-image is a ball lightning bolt. You never know where it's going to go. There are also collabs, where you can also use different neural networks, but I'm too lazy to explore all of it. So I have nothing to say about it using. I would like to get back to ArtBreeder in the future. There are a lot of interesting things I want to do, but I have only 24 hours in a day, so I have to prioritize so I don't get bogged down.
IGRUSHKINA: Oh, yes, I know how it is. I also want to learn a lot of things to try, and in the end there is no time to draw.
IGRUSHKINA: Okay. You think ArtBreeder is more interesting tool, but for some reason you still work with NightCafe? You can just switch from one to the other, you don't lose anything. And use a more advanced tool. In fact, you'll do the same thing - experiment and create art, only with more extensive functionality.
DEADWEIGHT: I don't think ArtBreeder is more interesting. I knew it was going to sound that way. It's just different, there are different mechanics. I only work at NightCafe now because I do work there for my "99 GANs" project. It's a conscious limitation. I want the 99 works in this collection to be done with one tool. It's like a series of works, a cycle. You determine how many works you want to do, you determine what tools they should be done with, and you don't deviate from those rules. Plus it's not distracting or entropic. I like to work within a very narrow framework. Or limited to mediums. In a narrow framework, the brain has to strain to produce new and new solutions, to get non-boring results. It's very stimulating.
"All Saints" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/506680
IGRUSHKINA: Good idea. Tell us more about your project 99 GANs.
DEADWEIGHT: GANs are generative adversarial networks. That's the acronym for this kind of neural networks that I use to generate these works. There is no essence, as in any creative act. There is a certain format that I come to. I create in it, I like the creative process itself. If someone likes what I get out of my creative process, that's great. It means it can be called art. If it's sold, that's even better, because then you can only do what you're passionate about and predisposed to. Everyone benefits from this: both the artist and the audience.
IGRUSHKINA: You could say that the point is within one series, to develop as an AI-artist, to do your own conscious exploration of working together with a particular neural nets to create a unique 99 work, focusing on working with one tool.
DEADWEIGHT: Yes. And another important part is working with blockchain technologies and the NFT ecosystem. Learning those mechanics, immersing yourself in the environment. Getting into the meta universe. That's VERY interesting, too.
IGRUSHKINA: There's basically - blockchain and the NFT ecosystem is just like one way to sell what you do...it's a marketplace. But going into meta universes is interesting. What do you think about that? How do we get into the metaverse with what you're doing now?
DEADWEIGHT: I disagree. Blockchain is as new and rapidly evolving as the field AI. I've been interested in it, too, since I learned about bitcoin in 2012. In that time, the blockchain technology sphere has progressed tremendously and is gradually becoming the new platform for the next stage of the Internet's development. This also includes the metaverse. In fact, we are already a bit in the metaverse. I have a virtual gallery with NFT works "99 GANs". One online museum room was not enough, I had to make three more to accommodate all 99 works. These four virtual spaces are connected to each other. The connection is made through a cryptocurrency wallet, any person from any country can enter. When pointing at any NFT picture you can go to its description, and further to the marketplace, where you can buy it. Having bought the work, it can be exhibited in another gallery. These mechanics of presenting your art and being able to sell it just didn't exist before this round of blockchain development. And these are solutions that are available for free. And this is just one aspect of the metaverse that is available right now.
"99 GANs" Drop 11/11- NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/542049
IGRUSHKINA: Yes, I understand, and I agree that blockchain is a developing and interesting theme. There is a lot going on in it. But I was asking in the context of the 99 GANs series. That's why I wrote that, in fact, blockchain and the NFT are only acting as a marketplace here.
DEADWEIGHT: Again, I disagree, because even the topics for 99 GANs, I choose in part to be related to blockchain and their memetics. So, blockchain here is as much an integral participant in both the creative process and the creative experiment. Its role cannot be reduced only to the marketplace environment. It is the third equal component in this cocktail.
If you dig deeper, I have several Telegram channels, I have regular playlists on Spotify, I have AI pictures presented in the form of NFTs. All of this can be connected together into a single mini-meta universe. The interaction and intersection of such micro universes of different authors and virtual worlds is essentially the metaverse. Under the hood are various decentralized blockchains. This enables the artist to interact directly with own audience. Both in terms of communication, interaction, and content creation and distribution. And also the ability to support your favorite artist directly, rather than buying a subscription from a centralized service, which will then transfer the author's honestly earned 5 cents.
IGRUSHKINA: I've always thought of the metaverse as a virtual space that you get into just when you put on VR glasses, and it's limited to what you see there. So if we're talking about creating something for the metaverse, it's 3D models that you can see in VR space. About the metaverse in the kind of context you're talking about, I haven't thought about it.
DEADWEIGHT: The 3D models are just visible part of the metaverse, yes.
"E-mail Pigeons" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/458932
IGRUSHKINA: The podcast that you sent earlier is on Spotify? Do you regularly host podcasts there? And, please, links to the Telegram channels in the studio!
DEADWEIGHT: The podcast is not on Spotify. I only have it on my Telegram channel and on Mixcloud. Made it a few years ago. it was interesting, but resource-intensive. After the first season, I suspended it, because I didn't find an audience. What is a podcast without an audience? I keep recording it now once a year, when I'm summing up the musical results of the year around New Year's Eve.
On Spotify I have 3 main playlists, mirroring the themes of my 3 telegram-channels: DEADWEIGHT LOGBOOK (multi-genre), GHETTO BLASTER (hip-hop) and MACHINESPEAK (electronic music). I fill up my playlists for a month, then start new ones.
https://t.me/weightdead is my main channel, I have 3 headings in it now: #trackweeklies, #trackmachines and album covers. Every weekend I download 8-12 music releases and listen to them for a week. I tweet the tracks I like right away and add them to playlists on Spotify. At the end of the week, I pick the best ones: #trackweek and #trackmachine (named after the on-board computer from the podcast). Sort of a weekly hit list. So it's not just in AI that I'm selecting, but in music as well.
https://t.me/urbanajungle - about hip-hop culture: rap and street art. I put rap music there and repost any content I like on the topic.
https://t.me/machinespeak - about technology and futurism. I throw electronic music there and repost important or interesting resources about technology. The other day I started doing a new column - interesting NFT artists. I find a lot of cool NFT artists now, I want to share beautiful things.
@DEADWEIGHT Spotify Podcasts https://open.spotify.com/user/ixairc0bmnfh3s55tvcvkh29l?si=c83ec2b1d82846cf&nd=1
IGRUSHKINA: Wouldn't you like to try experimenting with AI and the areas you used to work with - photography, video, music?
DEADWEIGHT: I'm already doing it a little bit. Making animations using AI. I really like working in that direction. I want to try video and I will definitely try it, but later. With music, less interest, but I have one eye on that as well.
IGRUSHKINA: How do you make animations with AI?
DEADWEIGHT: These are still short gifs from a set of iterations of one query.
IGRUSHKINA: What tools do you use to do this? I'm just into animation too...but frame by frame.
DEADWEIGHT: I take pictures from NightCAfe, and just collect them into gifs online.
IGRUSHKINA: Right now I'm having fun creating my first AI arts on your tip and I've got this idea. AI can also be a good tool for artists who have problems with imagination. You can work with AI and get stylistic ideas and images, which you can then use in your own work.
DEADWEIGHT: Yeah, some people also come up with this idea that you can use AI to generate concepts when your brain is tired.
IGRUSHKINA: Okay, about AI art. We've identified 2 formats: The first is when you combine one picture with another and you get the result. The second one is when you type in a word query and AI gives you the result.
DEADWEIGHT: There are many more. These are just the ones I've dealt with.
IGRUSHKINA: There's a popular thing in NFT now about working with neurons - generating characters randomly from elements that you give: different faces, skin colors, eyes, attributes, and so on. Have you tried doing something like that?
DEADWEIGHT: Personally, I am not interested in such experiments. Perhaps because I don't find the PFP (Profile Picture Project) format interesting. There's a whole slew of avatar projects. From CryptoPunks, and Bored Apes to Mutant Frogs.
"Fisherman" - NFT by @DEADWEIGHT https://objkt.com/asset/hicetnunc/483014
IGRUSHKINA: Last question for today. Let's talk about trends in AI. There was a thread on Telegram chat about some trends of AI working with pixels, for example.
DEADWEIGHT: Something new that came out is a pixel art generator, I don't know how to call it scientifically. It generates really cool pictures, like eye candy. It came out Stylegan3, but I have not seen anything interesting on it yet. Although the smoothness of its animation is amazing. From the very latest I've seen the generation of 3D models by text query. I think this is something that will shoot in the next phase. Who would deny themselves the pleasure of pulling this degenerate 3D crap into their metaverse?
DEADWEIGHT: Because AI often gives quite grotesque figures and there's a place to turn around in terms of creating all sorts of Frankensteins, if there's such an interest and inclination.
IGRUSHKINA: Pixel art generator - is it some kind of service like NightCafe? But only with pixels?
DEADWEIGHT: I don't know, I haven't check it.
IGRUSHKINA: Okay, I got you, now I have a lot of new topics to explore and find information. Thank you so much for agreeing to answer my questions. It's very interesting.
DEADWEIGHT: Thank you for such communication. It was interesting for me to answer your questions.
Thank you all for your attention. If you were interested or have something to add to the AI art theme, please just write in the comments about it. And maybe that will be a topic for the next conversation.
As a reminder, you can find all of @DEADWEIGHT's accounts at this link and learn more. https://linktr.ee/weightdead