The Spaceship Earth AI Project
Twenty Earth-Inspired AI Images That Should Completely Blow Your Mind
The Spaceship Earth AI Project is an art project composed of images developed using Midjourney. I need to briefly explain what “Midjourney” is for this article to make a bit more sense. Then we can talk about Spaceship Earth. Full disclosure, I have no financial or personal connection to Midjourney.
Midjourney Defined
At its heart, Midjourney is a computational engine — a computer program — that takes as input a prompt — basically a sentence or two of descriptive text and usually a few command switches or flags — and returns an image that is supposed to visually correlate to what was passed in. In effect, it is a computer program that generates Art from a text request.
This definition is wholly inadequate for a few reasons. For starters, it does not really communicate the experience of using the system.
There are several image generators available out there that by now most people would have heard of — DALL-E being the most recognizable name — but most of them operate in what I would call individual or private mode. No one participates in a DALL-E session other than the user, and no one can see what is going in, nor what is returned (other than, naturally, the operators running the service, Oz-like, hidden behind the curtain).
Discord Servers, Creating In Public
By contrast, Midjourney runs on a Discord Server as the user interface. What is a Discord Server? Basically it’s a chat forum or interactive bulletin board not so different from the chat rooms that used to be the crown jewels of America Online years ago. The way the Discord Server works is that people who want to use the image service sign up with the forum, are granted admission, have the rules explained to them, and eventually for a small sum gain access to chat rooms. Inside, commands are available to run the image generation software. People don’t actually do much chatting. But that does happen in the other forums, there is a help forum for example.
Midjourney is a lot of fun. It is interactive in the sense that once a prompt is entered and results returned, there are additional commands that can be applied to it, and the image more or less “grows” or develops under the guidance of different available commands. Mainly these are iterative.
It is all done within a space where other people can see both the inputs and the outputs. There are the stimuli of seeing what other people are doing going on somewhat confusedly within the same chat room, rapidly scrolling by. Also, perhaps a bit of intimidation at first about looking foolish and not knowing what to do. The “bot” that accepts the commands also sometimes communicates — maybe a warning about a banned word, maybe an indication that after a delay a requested job has started.
It’s a bit weird, too. But, for people like myself who routinely use Slack and Teams in the workplace and who spent a lot of time growing up in chat rooms this sort of thing is relatively easy to adapt to.
The Inspiration
Now, to the inspiration. A few weeks ago — yes, it’s that fast, a lot can happen in a few days with AI — I was in the chat called “general-4” which is where I tend to hang out, when I saw this prompt and the result:
planet mars with ice caps by Claude Monet
I unfortunately did not record the name of the user who originated this prompt. And to be clear, the above is not originally what I saw, I re-ran the prompt a few times for purposes of this article. Yes, to answer your question, unless a seed argument has been passed it is not possible to get back later the exact image with Midjourney, it is a proprietary system, and some aspects are opaque to the user. But the above is close to what I saw that night.
Now, these days I am just a guy who wants to retire by running an online store that sells t-shirts, but once I had some pretentions to be an artist. Well, at least I went to art school. That does not necessarily qualify me as an artist. I also had a whole career as a computer programmer. So, I have other things going on. Still, I am a guy on a mission to find new creative ideas and pretty good at seeing the potential in things. The above, “Mars as seen by Monet” woke something up in me. I almost instantly applied that to a related idea, to “the planet Earth as seen by Monet.”
Now we can take a small sidetrack to talk about the Whole Earth Catalog.
Stewart Brand/Whole Earth Again
As a child of a previous time I can speak eloquently about the impact of this big book, this catalog created by Stewart Brand which I suppose many of you reading this have never even seen. But, at the time, this was long before the internet — at the time, The Whole Earth Catalog was the hippie Bible of information, it was our World Wide Web. It had a big picture of Earth from Space on the front. And what inspired that book, its origin, was in an acid trip in which Brand thought he could perceive the curvature of the Earth. He connected his trip with the pictures coming back showing the Earth from deep space recorded by the astronauts of the Apollo Mission and understood, as perhaps the first person in history to understand fully, what those pictures meant to humanity.
You might like to read more about The Whole Earth Catalog.
Of course, today, Earth Day has become perfunctory and a corporate sell out, just as Greta Thunberg says. Green washing is the name of the game. But, back then, we really did get something of a high out of seeing that ghostly beautiful picture of the Earth floating in space. There was this children’s television program around that time — Big Blue Marble — that celebrated the vision.
The Spaceship Earth AI Project
So, flash forward to a few weeks ago, all that condensed in my head in a moment, and I could visualize it: imagine if the great painters and artists of times had access to that picture, what would it influence them to create? How would they interpret/interact with it?
This idea was heretofore impossible to provide an answer to, we could only speculate, and even then speculate wildly without facts. But there are two things which have come together to make it possible today, at least in a small way.
The first is the ability of the AI imaging systems to produce output that has many of the qualities of the art created by the great masters (or the lousy qualities of failed artists, it cuts both ways.) Form, content, themes, colors, everything you can imagine, the AI can recreate. It just must be asked for correctly. There’s a craft to that, just as there’s a craft to painting or photography or sculpture.
There are some caveats to this statement, of course. It all depends on the training set. In fact, neural networks are not really a new idea. They have been around almost as long as the computer itself, at least conceptually. But neural networks — pattern recognizers that are trained on images — with training sets of millions or billions of images from the public internet — that’s new. New, and powerful.
The second factor is the weird and remarkable ability of the AI to fuse different things together. This is a side effect of the specific algorithm used in the image generation process, called diffusion. What happens is the image starts out like a fog, and the biggest shapes and forms only are visible in the first few passes. Then as more passes are made over the image, details are evolved into view. That’s diffusion. I am not explaining it in a precise way. Here’s what that looks like with another AI system, one that can be more easily controlled at home, called Stable Diffusion, at 2, 6, 10, and 20 steps in the diffusion process:
The prompt here was “earth as seen from space.”
This aspect of diffusion is interesting, but not really that strange because it is more or less how a human painter might construct a canvas by making passes over it from large and abstract — cones, sphere and cubes — to specific details of plants, faces, or buildings.
However, it gets weirder. And this brings us to the Spaceship Earth AI Project.
Here’s what happens when you ask for Leonardo Da Vinci’s impression for Earth as Seen From Space:
Prompt: “Ink Drawing on parchment of planet Earth as seen from space. We can see ice caps and continents, small city lights. symbols of mystery and small drawings of figures, machines, and plants. style of Leonardo Da Vinci, ink on parchment, magical, insane detail and intricacy.”
Four outputs using different seeds:
Whoa! Looks very “Leonardo” doesn’t it?
Here’s a look at a later evolution of the top left image, which is achieved through upscaling:
You can clearly see notes in the style of Leonardo drawings — his notebooks were full of drawings with copious notes — as well as a central core of roundish objects held together with a sort of wire frame.
OK, so this is not a globe so much as the inner workings of an organic structure. Perhaps this is how Leonardo the artist/thinker conceived and imagined his world? Did we just learn/discover something astonishing?
Now, you will agree that my prompt didn’t indicate anything specific. I asked for “symbols of mystery and small drawings of figures, machines, and plants. style of Leonardo Da Vinci, ink on parchment, magical, insane detail and intricacy,” and I got that to some extent. But what I got is also very much “Leonardo” the artist and not what was asked for. To my eyes it looks a bit like the famous embryo studies page from the notebooks showing the inside structure of a womb and the nascent being within:
Maybe Ihaven’t impressed you yet, let’s look at another example from the Spaceship Earth AI Project that shows the subtle fusing of the prompt with pre-existing and external knowledge present in the Midjourney AI training model. The prompt here is:
“A Medieval manuscript pen and ink drawing of planet Earth as seen from space. style of the Lindisfarne Gospels. We can see ice caps and continents, small cities and signs of inhabitation. magical, insane detail and intricacy.”
Output:
I think you’ll agree there’s nothing in my prompt to suggest the Earth is flat, although this was believed widely in 715 AD Europe. And yet the AI algorithm made a facsimile of a page in unintelligible Latin with the suggestion of a small disk partitioned into longitude and latitude lines and a moon, but it also produced an island universe, a flat earth floating in a sea of space, as an illustration for the page.
So this kind of thing is both spooky and evocative. It is a reminder to all the people who do not believe this is “Art” as of yet; that within the diffusion technique and a well-trained model, there are things that artists formerly would have ascribed to the Collective Unconscious, and who are themselves somewhat taken aback by, when those appear in their work.
The Spaceship Earth AI Project, which is essentially a business venture by me intended to put interesting art about the Earth as seen from Space on T-Shirts for people to buy — also has intrinsic artistic value. We are exploring a new kind of art, something that to be honest only a Steward Brand or someone influenced by that optimistic line of thinking would perceive. But it’s there. Let’s look at a few more examples.
The Proof of Concept: Paul Klee
Paul Klee was a Swiss-born German artist who lived in the first part of the 20th century. His highly individual style was influenced by movements in art that included expressionism, cubism, and surrealism. He’s one of my favorites due to the sensitivity, color, and tactile quality of his visual works. Here’s my “Earth as seen from Space” prompt for Klee:
Prompt: “Painting of planet Earth as seen from space. We can see ice caps and continents, small cities and signs of inhabitation. Paul Klee. magical, insane detail and intricacy.”
Output:
When I saw this for the first time I realized I was hyper-ventilating and had to stop and take a break. My notes indicate it was only the second prompt after receiving the original inspiration from the “Mars by Monet” output. This was just as I was trying to determine in my own mind if the idea was good. See, it’s easy for an artist to have ideas. That doesn’t mean they are all good ideas!
But upscaling produced this:
What can we say about this? Well, it’s pure Paul Klee style, right down to the textures and colours in a grid — one of his signature design ideas. You can see this grid idea in works like Flora on Sand, 1927:
Also interesting is the AI’s selection of a picture fame right down to the museum show label at bottom left. Clearly there were some museum images in the public Internet training set.
But the point here is that I didn’t ask for a grid of colours. The AI provided that. Nor is there anything precisely like this in all of Paul Klee’s works, at least that I am aware of.
So this is the effect of the diffusion technique, which can seamlessly meld together different images in a way that is similar to compositing, but also different due to the level of control. It’s not about compositing at the level of pixels, of raw images. It is something deeper at work at the level of ideas and concepts.
But Is It Art?
Is it art? Of course it’s art. Now full disclosure, not everything output on Midjourney is good. It’s all art if the person using the AI is attempting to make or obtain an image that correlate to their prompt, AND the prompt correspondingly attempts to say something meaningful. This is a bigger discussion I will attempt at another time. But for now, accept the obvious evidence of your eyes that this is art.
What went through my own mind was, good Lord, is this commercially viable work? Let’s look at a few more of the Spaceship Earth AI Project artifacts and you can share your thoughts about that with me in the comments.
The rest of this article shows some of the outputs from the project to date. Enjoy!
Rembrandt
Rembrandt Harmenszoon van Rijn ( 15 July 1606–4 October 1669), usually simply known as Rembrandt, was a Dutch Golden Age painter, printmaker and draughtsman. An innovative and prolific master in three media, he is generally considered one of the greatest visual artists in the history of art and the most important in Dutch art history. Wikipedia.
Prompt that did not produce useful results: “Painting of planet Earth as seen from space. We can see ice caps and continents, small city lights and signs of inhabitation. Rembrandt, old master, brushstrokes, magical, insane detail and intricacy.”
Final Prompt: “A self-portrait of Rembrandt holding a globe of the planet Earth as seen from space. Style of Rembrandt Harmenszoon van Rijn. magical, insane detail and intricacy.”
Output:
I was not able to produce usable results with the outward-looking and world-expanding “view of planet earth as seen from space.” For Rembrandt, and some others, this simply did not compute. But what did work, and I think thematically is correct, is to inject the guidance “holding a globe of the planet Earth as seen from space” and then things came together.
I think we are on firm ground to say there is no self-portrait by historical Rembrandt that contains a globe. I hope those who know about these things will correct me if I am wrong. I will draw your attention to one detail in this remarkable image, the shadow that extends outside of the portrait frame. The old masters were interested in tricks to make their work more realistic. It was something that buyers — the customers — found intriguing. They craved what we might call hyper-realism, in a time 200 years before photography. So, this shadow over the frame achieves hyper-realism. Maybe it sounds remarkable for an AI to pick that up. You be the judge.
About his hand. This was not originally rendered in a pleasing way by Midjourney, and it was necessary to resort to another engine — DALL-E — to do inpainting on the hand. I doubt if I had not mentioned it if anyone would have noticed the slight visual difference in colour, but at least I know. These are the sorts of things the developers of AI are still working through and that artists working with the system learn about.
For reasons that I don’t understand, DALL-E can render perfectly acceptable hands most of the time (say 50% usable). That doesn’t sound great, but it is compared to the other engines. Midjourney is not good at all with hands. To be fair, human hands are weird looking generally, taken as pure forms, and are very hard to render in 2D anyway (we as homo sapiens don’t necessarily know this about ourselves, but the AI does). I could write a whole Medium.com post about this, and how AI can teach us things about ourselves that we don’t know, but we will save it for another time.
Edgar Degas
Edgar Degas, born Hilaire-Germain-Edgar De Gas (19 July 1834–27 September 1917) was a French Impressionist artist famous for his pastel drawings and oil paintings. Wikipedia.
Prompt: “A painting of a dancer holding the planet Earth. Impressionist style of Edgar Degas. rule of thirds, magical, insanely detailed and intricate.”
Output:
This is one of my favourite images from the whole project. She’s not “holding the earth” but she holds the picture together as the centre of attention behind the luminous globe. You can see how the greens of the globe — instead of the expected blue marble — merge with the background colours and complement the browns of the dancer and the floor. An almost unimaginably exquisite Degas.
Paul Gauguin
Eugène Henri Paul Gauguin (7 June 1848–8 May 1903) was a French Post-Impressionist artist. Unappreciated until after his death, Gauguin is now recognized for his experimental use of color and Synthetist style that were distinct from Impressionism. Toward the end of his life, he spent ten years in French Polynesia. The paintings from this time depict people or landscapes from that region. Wikipedia
Prompt: “A painting of planet Earth as seen from space, post-Impressionist style of Eugène Henri Paul Gauguin. Tahiti, Tahitian girls, “what are we here for, where are we going”, totem idol, rule of thirds, magical, insanely detailed and intricate.”
Output:
It’s so easy to get good effects with Gauguin prompts. The reasons are not entirely clear, perhaps the intense colours are beguiling and the tropical settings naturally inviting.
Katsushika Hosukai
Katsushika Hokusai (葛飾 北斎, c. 31 October 1760–10 May 1849), known simply as Hokusai, was a Japanese ukiyo-e artist of the Edo period, active as a painter and printmaker. Wikipedia
Prompt: “A painting of planet Earth as seen from space, style by Katsushika Hokusai, 葛飾 北斎, woodblock printing from Japan. “The Great Wave off Kanagawa”, rule of thirds, magical, insanely detailed and intricate.”
Output:
Too easy. But interesting none the less.
Surrealism/Rene Magritte
René François Ghislain Magritte was a Belgian surrealist artist, who became well known for creating a number of witty and thought-provoking images. Often depicting ordinary objects in an unusual context, his work is known for challenging observers’ preconditioned perceptions of reality. Wikipedia
Prompt: “A painting of the planet earth broken into pieces and reassembled, Surrealism, Dada. style of René François Ghislain Magritte, rule of thirds, dream-like, magical, insanely detailed and intricate.”
Output:
I would draw your attention here to the fact that the requested “earth” looks a lot like a cantaloupe (musk melon if you live in certain parts of the world). Also that there is a dark mark, that looks like a divot or depression, in which the melon would fit, and that in terms of composition this gives a sense that the melon-earth will fall, or at least that it has the ability of motion and could come down at any moment. There are proto-earth-melons developing in the clouds, off in the distance. And finally, some of the hedges to the sides of the composition have a distinct animal quality, something like elephants, recalling the “hills like white elephants” of Hemingway.
This is first-rate Surrealism.
Pablo Picasso and Georges Braque/Cubism
Picasso is almost too easy with AI, and the kind of images generated do not interest me all that much because there is a certain element that is lacking. But it is more an issue of how the prompt is formulated. For those types of art that have been over-exposed in the cultural zeitgeist, that are almost caricatures of themselves by now, a certain amount of effort is required to get something new out of it. For Spaceship Earth, I needed to add “portrait of a female” into the mix to make it output the kind of results I wanted.
Prompt: “Portrait of a female planet Earth as seen from space. cubist style of Pablo Picasso. abstract. rule of thirds, magical, insanely detailed and intricate.”
Outputs:
I don’t know if you can see at this resolution the detail of the eyes, but they are very painterly and expressive — almost like drawn from life, or drawn by an exquisite draftsman, which Picasso was. The Harlequin element is also in evidence with the Earth as a background disk and echoed in the shapes.
With Georges Braque, a similar prompt was also useful:
Prompt: “Portrait of a female planet Earth as seen from space. style of Georges Braque. abstract. cubism. rule of thirds, magical, insanely detailed and intricate.”
Output:
For Braque, the AI has produced a somewhat dour portrait of a cold beauty, with alive blue earth on the left and parallel dark, dead planet on the right. There is a similarity to this and the above prompt that eerily feels like the same model sat for both portraits.
Soviet Realism
Socialist realism is a style of idealized realistic art that was developed in the Soviet Union and was the official style in that country between 1932 and 1988, as well as in other socialist countries after World War II. Wikipedia
Prompt: “A painting of communist party workers looking at a globe of planet Earth, style of Soviet Realism. propaganda, rule of thirds, magical, insanely detailed and intricate.”
Output:
I did not spend much time on these and they look less good at higher resolution, but the thing that stands out for me is the correctly imagined grotesque vision of the Soviet and Communist Chinese State in relation to their view of the Earth and the human beings on it. The Earth as a stone core, or a reflecting ball for a few party leaders — that sounds about right.
Art Deco/George Barbier
George Barbier, né Georges Augustin Barbier, (1882–1932) was one of the great French illustrators of the early 20th century. Born in Nantes, France on 16 October 1882, Barbier was 29 years old when he mounted his first exhibition in 1911 and was subsequently swept to the forefront of his profession with commissions to design theatre and ballet costumes, to illustrate books, and to produce haute couture fashion illustrations. Wikipedia.
Prompt: “A painting of planet Earth as seen from space, Art Deco style of George Barbier. rule of thirds, magical, insanely detailed and intricate.”
Output:
This beautiful egg-like form splitting the earth and sky in a 1:3 proportion is unique in all the images I have created so far with similar prompts. Looking through Art Deco images I have not found its like. Surely the AI did not create something new? Of course not. How could it? Right?
Animation/Chuck Jones
Charles Martin Jones (September 21, 1912 — February 22, 2002) was an American animator, director, and painter, best known for his work with Warner Bros. Cartoons on the Looney Tunes and Merrie Melodies series of shorts. He wrote, produced, and/or directed many classic Animated Cartoon shorts starring Bugs Bunny, Daffy Duck, Wile E. Coyote and the Road Runner, Pepé Le Pew, and Porky Pig, among others. Wikipedia
Prompt: A cartoon cell depicting the earth from space and astronauts, style of Chuck Jones, Looney Tunes, rule of thirds, gritty, magical, insanely detailed and intricate.
Output:
I love this image because it realistically paints the earth as might be imagined in the mind of a Looney Tunes cartoon — something I grew up on but younger people may be less familiar with. But the style is there of the Wily Coyote Roadrunner Hills and deserts. Notice that there are a few small characters in the frame but nothing that would draw the ire of the corporate intellectual property fun police.
Akira (1988)/Japanese Animation
Akira (Japanese: アキラ) is a 1988 Japanese animated cyberpunk action film directed by Katsuhiro Otomo, produced by Ryōhei Suzuki and Shunzō Katō, and written by Otomo and Izo Hashimoto, based on Otomo’s 1982 manga of the same name. Set in a dystopian 2019, it tells the story of Shōtarō Kaneda, a leader of a biker gang whose childhood friend, Tetsuo Shima, acquires incredible telekinetic abilities after a motorcycle accident, eventually threatening an entire military complex amid chaos and rebellion in the sprawling futuristic metropolis of Neo-Tokyo. Wikipedia
Prompt: “A cartoon cell depicting the earth from space and astronauts, style of the film Akira, Katsuhiro Otomo, manga, Neo-Tokyo, rule of thirds, gritty, magical, insanely detailed and intricate.”
Output:
We note the request for Neo-Tokyo graphics is nowhere near as intense as what goes on in the movie but has a cool blue cityscape with a huge red blight, like a cancer, growing on the side of the world and covering much of it. The blight is as big as a continent. Recall that thematically Akira contains a story line where one of the characters morphs into a gigantic ball of flesh that devours everything in its path. The image has red light effects that are also weak but at least reminiscent of the light effects in the original film.
My point here is that the AI picked up on the thematic element present in Akira and provided something that contains more content than one might have anticipated from the prompt. A little spooky in that sense.
Judy Chicago
Judy Chicago (born Judith Sylvia Cohen; July 20, 1939) is an American feminist artist, art educator, and writer known for her large collaborative art installation pieces about birth and creation images, which examine the role of women in history and culture. During the 1970s, Chicago founded the first feminist art program in the United States at California State University Fresno (formerly Fresno State College) and acted as a catalyst for feminist art and art education. Wikipedia
Prompt: “A painting of the feminine, banned from history version of the planet earth as seen from space, Feminist, style of Judy Chicago, rule of thirds, flower, magical, insanely detailed and intricate.”
Output:
This clearly captures something of the first-wave feminism from the famous Dinner Party work. Notice that the image still melds with that of the Earth and that there is a transition from the flower vagina motif to the earth disk — a tan ring of colour that perhaps represents soil, then atmosphere, then finally the darkness of space. At top right we can see a bit of landscape that the flower petals transition nicely into as mountainside. The image is integrated and has flow, both chromatically and as forms — it is not a raw composite or a rip off.
Futura 2000/Graffiti Styles
Leonard Hilton McGurr (born November 17, 1955), known as Futura, and formerly known as Futura 2000, is an American graffiti artist. Wikipedia
Prompt: “A photograph of graffiti depicting the earth from space and astronauts, style of Futura 2000, spray paint on a brick wall, rule of thirds, gritty, magical, insanely detailed and intricate.”
Output:
It was surprisingly difficult to get a good result for Futura, considering the size of his visual record, and for graffiti generally. The idea that I had in my mind for this image was different, and this is not sufficiently spray paint-like to my eyes to be realistically assigned as spray paint media. Perhaps it could be achieved with an air brush. Still, it is an interesting image on a wall, and rendered with a photographic or lens eye. The Earth is melting, dripping vermillion. Well, perhaps that is saying something to us.
Banksy
Banksy is a pseudonymous England-based Street artist, political activist and film director whose real name and identity remain unconfirmed and the subject of speculation. Active since the 1990s, his satirical street art and subversive epigrams combine dark humour with graffiti executed in a distinctive stencilling technique. Wikipedia.
Prompt: “A photograph of graffiti depicting the earth from space as a balloon, style of Banksy, rule of thirds, gritty, magical, insanely detailed and intricate.”
Output:
I’m not sure if it is clear from the image at this resolution, but what has been rendered is a picture drawn or stencilled on a wall. That’s one of the things that makes this so remarkable. It may be that asking for a photograph causes this effect or that Banksy art is always on walls.
Compared to the difficulties I had with Futura, Banksy is actually quite amenable to AI generation. It may be that his style is flatter or somehow easier, I cannot really explain it. For evidence, here is another recent “Banksy” not related to the Spaceship Earth AI Project.
Midjourney As Itself
It turns out that Midjourney itself has a style — a distinct way things look if you do not provide much guidance. Just as a bonus image, here’s what that looks like:
Conclusion
I have more to say about AI, but your minds really should already be completely blown at this point. If you want to buy any of the art you see above (which is all copyrighted by the way) that is possible to achieve though my online store. Not everything is up yet, I have been busy with image generation but it will all be up there soon.