The Best AI Art in the World, Vol. 7

This new Retro series—curated by a longtime professional cultural critic and former Digital Arts professor at University of New Hampshire—highlights amazing Artificial Intelligence-generated art.

ABOVE: “Picasso Batman.”

---

Get a free trial of Retro below! Enjoy a dozen sections and 200+ reports and cancel anytime.

---

{Note: This Retro series features ten works of AI art in each volume—in some cases, series of such artworks—offered in no particular order. To get a sense of the perspective of the curator of this collection with respect to AI art, see this essay. The artwork names that appear below are those that were attached to the work upon its publication, or, if no such title was provided, an approximation of the prompt used to create the art as announced by its “algo jockey”/“AJ”: the individual who crafted the text prompt that produced the artwork. The images below hail from major AI-art apps such as Midjourney, DALL-E 2 (built by OpenAI), Craiyon, Dream (built by WOMBO), or Stable Diffusion (built by Stability AI). The source for each artwork, if known, is linked to immediately following each artwork or series of artworks. AJs who would prefer not to have their work appear in this series can request the removal of a work via DM.}

---

Introduction

All of the artworks that appear below were created by artificial intelligence (“AI”).

If you would like to create some AI art yourself, check out this informative thread.

The works that appear below are not copyrighted, are not reproductions of existing works, and were produced by artificial-intelligence protocols that devised them after being “trained” on billions of publicly available images online. A prompt similar to the one used to generate the images below, reentered into the same protocol, would likely result in a different work—as would entering the same prompt into a different AI protocol (see the note atop this article for a list of some of the popular AI-art apps you will see represented as this Retro series unfolds in the coming months and years).

Except where otherwise indicated, the pictures below are not intended to accurately depict any living or historical person or brand. Any resemblance of human figures to individuals presently alive is—again, except where otherwise indicated—coincidental.

None of these images have been digitally altered from their original publication, with the infrequent exception of cropping to elide (a) inadvertent bordering, or (b) branding that distracts from the artwork and is unnecessary given that the source of each work (as noted in the italicized text above) is linked to below each work or series of works.

In those instances in which Retro offers a “series” of thematically and/or substantively linked AI-generated artworks, please note that the series has been curated, meaning that not all of the artworks originally appearing in the series have been included here. As the procedures undergirding this series involve using curatorial journalism (and my years of experience as an art critic and digital studies professor) to offer readers the best exemplars of AI art available in the estimation of this author, and because AI-art apps will for some time to come occasionally produce substandard artifacts in reply to a given prompt, this additional layer of curation is necessary to maintain the high standards of both this series and Retro writ large.

The overarching purpose of this new Retro series is not just to demonstrate the quite startling range of contemporary AI-art apps’ generative capacity—that is, to celebrate and honor the works that appear in it by attaching to them the breathless superlative that gives this series its name—but also to offer a quick-link to the “algo jockey” who crafted the input that produced each image. As noted above, if any such “AJ” wishes to have Retro take down a particular work, Retro is happy to do so and do so promptly.

The Evolution of AI Art

One thing readers must keep in mind as they follow this series is that AI art is getting better by the day—quite literally. Retro expects that as the major AI-art apps continue training themselves on publicly available images and the thousands and thousands of new text prompts they’re getting daily, we will not only see algo jockeys becoming more and more talented at what they do (and there are already some clear superstars in the space, whose oeuvre instantly ends the notion anyone can produce superlative AI art without a deep understanding of the tools now available for making it) but we will also start to see these apps’ protocols learn how to better respond to text prompts.

So while it’s tempting to treat AI art as a static concept, we must not. It’s a moving target that may well look very, very different in 2024 and 2025 than it did in early 2023.

What It Takes to Become a Superlative AI Artist

Those who have spent no time in the world of AI art, and/or no time around AI artists, are apt to think of AI art production as a simple process that even a fool could execute—and in a limited sense, that’s true. Just as a child, given a set of paints, could throw those paints around a room and, in getting a bit on the walls here and there, create a accidental, vague, tragicomic homage to Jackson Pollock, anyone who uses (say) Bing Image Creator will in short order produce a piece of AI art. It will almost certainly be quite a bad piece of AI art, but it will be AI art nonetheless. In the same way one can “play” baseball simply by standing at home plate with a bat and facing the pitcher; in the same way one can be a “photographer” by just pressing a button to take a photo; in the same way someone bopping around their bedroom in the dark can in a general sense be said to be “dancing,” anyone who wants to generate AI art can. Just not well.

The best AI artists tend to have a background in painting, photography, or both. This enables them to understand the basic precepts of visual representation that must be considered in generating a successful text prompt and then minutely editing it over dozens of rounds of drafting; if you haven’t told the AI to consider color, form, line, shape, space, texture, value, resolution, angle, perspective, dimension, and the scores of other fundamental terms in the visual arts that visual artists are attendant to, the AI is not going to consider those elements and thus your work will lack any abiding attention to them.

By the same token, if you lack the experience or taste to know what good photography and painting look like—which requires the ability to study an artwork and see what (if anything) it is missing, only poorly includes, or altogether fails to be—you won’t know how to curate the scores of variations on your text prompt that the AI gives you. Even a perfect prompt will fail if the AI artist doesn’t know which exemplars of the artwork generated by that prompt are fit for public consumption. And many artists certainly don’t know this (as is also true for many whose art depends on their own motor skills).

Just so, a superlative AI artist knows their tools every bit as obsessively as a graphic designer knows her applications or a painter her paint and brushes. Already, the best AI artists are giving public lectures and attending symposia to understand the many complex terms and subprocesses AI-art apps allow users to employ to create the best artwork possible. Those who just click around on Midjourney for a few minutes will by no means understand image blending, image description, “knolling”, and so on to the degree necessary to create a superlative AI artwork. And here’s a further rub: the number of such terms and concepts is growing by the week for most of these AI-art apps, as is the ability of the most superlative users of these apps to tweak their text prompts to squeeze the most out of the UI they’re working with. So it’s getting harder and harder, practically by the day, to distinguish oneself as an AI artist. It already takes hundreds of hours of trial-and-error (and studying the discoveries others have stumbled across and shared) to gain real facility with even one AI-art app, let alone all several dozen of them.

And this, finally, is the point so many seem to miss—and the reason there will never be any more “great” AI artists (as a percentage of all artists) than there currently are great photographers, great dancers, or great painters. The simple fact is that almost no one is going to want to dedicate the thousands of hours required to become among the best AI artists out there. Great art, whether AI-aided or not, requires an obsessive compulsion to learn how to wring every last drop of capacity out of a given tool, which compulsion must also be preceded by enough conventional knowledge of the visual arts to know what questions one must ask oneself in striving to gain expertise with a given tool. In other words, time alone doesn’t generate expertise; guidance in how to spend one’s training time is also required, which is particularly problematic in the context of AI art production. Do not doubt that there will eventually be semester-long university courses in how to become a commercially viable, artistically superlative AI artist, just as there are already such courses for graphic design or (for that matter) any subgenre of the visual arts that is predominantly dependent on an individual’s fine motor skills.

An argument can even be made that AI art production is harder than some other forms of art production because there are, as yet, few professional guides for those seeking to learn to do it well and (for now) no dedicated academic courses. A hundred percent of the best AI artists working right now are nearly 100% self-taught. And even the very strongest AI artists alive today are still learning their craft in real time because their craft is only a matter of months (at best a few years) old and volatile to a degree no far older subgenre of the visual arts is; compare this to painting, say, or even dance, which have been extant in human culture for millennia and therefore—while extremely difficult to do well and admittedly still evolving—forms of expression that can at least offer their aspirants a thousand thoughtful road maps for how to climb that particular mountain.

With AI art, most of the roads to greatness remain untraveled and some have not yet even been imagined.

Retro has been approached by many people who dislike AI art without knowing much about it. The general refrain heard from such critics is that AI art can’t be art because it requires no skill. Of course it really requires enormous skill—just not fine motor skills.

It requires writing skills, research skills, curation, a knowledge of aesthetics, a great deal of technical expertise, taste, dedication, perseverance, self-critique, and so many of the components that all art instruction (and I say this as someone who taught the arts at the university level for years) necessarily incorporates. Other critics of AI art make the stunning claim that AI art requires no imagination; while this canard hardly deserves a response, suffice to say that AI artists—just like all other artists—form in their mind’s eye a composition they wish to see in the world and then expertly use the tools at their disposal to make what they have imagined a reality. Indeed, an AI artist having less direct control over her tools than (say) a painter with her many expensive brushes only makes it harder for an AI artist to manifest what is in her imagination in digital form. Finally—and putting aside those who simply say AI art is not art because anyone can do it, a fallacy addressed earlier in this essay—there are those who say that AI art lacks the human and emotional components of the various visual, material, and performing arts. Tell this to sculptors who work in the digital realm, or professional storytellers, or those who can sing but can’t play any externalized instrument. Tell this to graphic designers, or to architects who hire construction crews to build what they imagine. AI art does not exist without a human vision or human input, and the human vision in question can’t exist without the same emotional impulse to create that artists of every stripe finally must have.

AI art is in its first months and years. It will face the same preposterous prejudices that free-verse poetry faced in the mid-nineteenth century, when Walt Whitman—now considered, with Emily Dickinson, one of American poetry’s parents—was told that Leaves of Grass “wasn’t poetry” because it lacked rhyme and meter. More recently, conceptual poets have been told conceptual poetry “isn’t poetry” because it doesn’t often rely on image, metaphor, or the condensation of language, and in some cases (in an echo of AI art) even uses exclusively words first devised by others, though their novel arrangement is entirely the work of the conceptual poet and the quality of the final work is dependent on that poet’s own ingenuity, judgment, curation, and taste.

There is no art form that emerges without its detractors, and no detractors of an art form still early in its development act as responsibly as they would if they were instead applying their critiques to artworks already universally accepted as artistic in character. Because there’s no natural constituency for AI art besides those who make it—at least not yet—there is little impetus for the critics of such art to be fair to it and indeed much encouragement for them to “take a stand” by being categorically against it. Certainly, we have seen online how conventional painters and photographers have cheered even the most factually incorrect and deeply disingenuous critiques of AI art for the simple reasons (if also understandable ones) that (a) conventional artists worry their future earnings as freelance artists could be impacted by the availability of AI-art apps, and/or (b) they see that those who oppose AI art publicly get feted for doing so by their peers—which in the art world can be a professionally wise space to occupy.

Certain critics of AI art, making certain criticisms of AI art, may be right to worry, as discussed here. That does not mean their objections are accurate, made in good faith, or worth being accepted at face value.

The upshot is this: AI art is getting better by the day because the very best AI artists are getting better by the day; everyone else—both critics who fail to investigate what AI art actually is and how it works, or lovers of AI art who nevertheless refuse to put the time and effort into it that it requires—is falling farther and farther behind by the hour. This Retro series aims to remain as best it can at the bleeding edge of AI art, always celebrating the diversity and wonder and craft of this emerging art form even as it understands, too, that the animus against it will continue for many years to come and in certain instances will be warranted.

{Note: On rare occasion, images that appear in this series may be unsettling to certain viewers. Any such upset is unintentional; these artworks are intended to surprise, delight, illuminate and amaze, not provoke negative emotions or physical responses. However, because any AI-generated product is definitionally already lodged deep in the “uncanny valley”, occasional strong reactions to the images in this series are to be expected. Certain images are prefaced by a “curator’s note” that briefly explains why I find this particular work to be of artistic merit.}

---

(1) “Tiny” (2023) [Series]

[Source.]

---

(2) “Midjourney Style Tour” (2023) [Series]

Curator’s note: Most of these images are wondrous, but be aware that some are mildly unsettling—even as they are all superlative exemplars of advanced AI art-generation techniques.

Ancient Metalwork

Stippling

Patchwork

Action Figure

Chalk Art

Polaroid

Mokume-gane

Diagrams

Metallic Foil

Bone Art

Tissue Paper

Optical Illusion

Felt

Chiascuro

Anaglyph

Carbon Dust

Pop Art

Postage Stamps

Tapestry

Vignettes

Kintsugi

[Source.]

---

(3) Three Series (2023)

“The Valley”

“City at Dusk After Rain”

“Liquid Colors”

[Source.]

---

(4) “Batman, As Painted By Famous Painters” (2023) [Series]

Curator’s note: Normally I “curate out” any “four-square variations” to ensure that only the best work is highlighted. I have declined to do so here because I think the project below is as much about art education as anything else. In disclosing several variations on a single theme from a number of famous painters, this series gives us a sense of the many idiosyncratic qualities of each artist that the AI is attuning itself to. It takes seeing all the artworks in this series to fully appreciate both these artists and their aesthetics, not to mention—of course—the unique opportunity we now have to see these artists’ talents applied to an iconic superhero most of these artists had never heard of.

Sorolla

Basquiat

Velazquez

Pollock

Da Vinci

Dali

Vermeer

Monet

Rembrandt

Magritte

Klimt

Goya

Warhol

Caravaggio

Bosch

Cézanne

Rockwell

Gauguin

Seurat

Wood

Picasso

Banksy

Van Gogh

[Source.]

---

(5) “Lady Deadpool, Flapper” (2023)

[Source.]

---

(6) “Unseen Meme Angles“ (2023) [Series]

Curator’s note: This series takes advantage of the wonderful “generative fill” function of the latest Photoshop update. Only the content inside the white rectangle is from a source originating outside Photoshop. (I realize that readers may have to read that last sentence a few times!)

[Source.]

---

(7) “Extreme Influencer Selfies” (2023) [Series]

[Source.]

---

(8) The Daily Life of a Terracotta Warrior” (2023) [Series]

[Source.]

---

(9) “Solarpunk” (2023) [XAJ Series]

Curator’s note: “XAJ” indicates that “multiple algorithm jockeys” were involved in the creation of a series. Retro strongly encourages and welcomes such collaborations. For the series below, all entries’ prompts were inspired by (and also included) the term “solarpunk.”

[Source.]

---

(10) “The City Beyond the Forest” (2023)

[Source.]

---