Tim Arends has a Bachelor of Arts degree in Fine Art from Berea College, KY. He has written on a variety of technological subjects.
DALL-E 2: The Artificially Intelligent Artist
Perhaps you haven’t heard yet about the amazing, artificially intelligent artist called DALL-E 2, the AI that can create fantastic drawings and paintings for you—from scratch!
No, this is not a Photoshop or Instagram filter that merely alters a photo that you give it. This is an AI that creates fantastic and beautiful images from a blank virtual canvas.
And here’s the most amazing part: You don’t have to be an expert to use it. You simply tell it what you want, as if you were talking to another human being, and it gives it to you. And its work looks like it was done by a talented professional artist with years of experience and training!
Look at the image above. Observe the lighting. Take in the cozy ambience. Look at the drinks for two on the table. Notice how you can practically feel the texture of green paint on the wall.
No, this is not a photograph—it was "painted" by DALL-E 2 simply by being given the text prompt above. It is a unique piece of art. It was not copied from photographs that already exist. There is no other image quite like it in the entire world.
The upshot of this is that not only can DALL-E 2 create fantastic art, but the AI understands what you are saying to it as well or better than a real person does! It is able to intuit from a simple text prompt the emotional tone you are attempting to imbue into an image, and it gives it to you!
It truly is like magic. It’s like more than magic, it’s like DALL-E 2 is alive. A sentient being. As creative, understanding and intuitive as you and I.
But, as amazing as this technology is, its arrival raises a whole host of questions.
AI Art Raises Questions
The act of creativity is considered the quintessence of being human. Art is considered an expression of the human soul. It is the way we communicate with others. It is a visual expression of our emotions—our very being.
So what does it say when a machine is able to create artwork that is indistinguishable from that created by humans—art that contains emotion, passion, drama, a sense of composition and beauty, even a sense of humor?
What will the arrival of AI artists do to the professional art field? Who will be considered the creators of AI-generated artwork? Will artificial intelligence soon take over all jobs? Is there any point to drawing and painting anymore if an AI can do it as well or better than we can?
These are just a few of the thorny questions we will tackle in this essay.
Will AI Artists Like Dall-E 2 Take Jobs From Human Artists?
Yes, this is almost certain to happen, but I do not think it will destroy the jobs of all artists, nor do I think it will happen overnight. It will definitely be a gradual process.
There will always be human artists, but it will become increasingly difficult to make a living in many fields of art. Many established artists will get by on the strength of their reputation and the following they have managed to generate over the years. Caricature artists of the type you see at fairs, festivals and private parties will continue to be popular on the basis of the strength of their personality. It’s a lot more fun and engaging to be drawn by a real person than by an iPhone app!
AI generated art is not likely to hang on the walls of art museums (unless it is part of a special exhibit of computer art). But in the fields of both commercial and fine art, the supply has always exceeded the demand and that is likely to become even more true in the age of AI artists.
Advertising agencies, for example, may question why they should hire a human artist to create a drawing of a boy holding their product when it’s so quick and easy to have an AI artist create the same image. Even photographers may be in real danger if an AI can create a photograph-like image of a girl holding an umbrella in the rain, eliminating the need to hire a photographer to capture the same image.
I think [DALL-E] will really threaten photography both as a hobby and as a profession.
— Micael Widell, Photographer (YouTube)
Will AI Art Render Human Creativity Pointless?
Some people may ask: Why bother doing art even on an amateur basis when you’ll never be as good as the machine anyway? But the same thing can be said about playing chess. Computers now routinely beat grandmaster-level chess champions, and yet the hobby of chess is as popular as ever. Just because a computer can play chess as well or better than most humans doesn’t mean there isn’t an intellectual satisfaction to the challenge of playing chess.
The same is true of trivia games. Even though IBM Watson became the champion of the TV game show Jeopardy in one episode, that didn’t cause the show to go off the air. Likewise, even though a motorized vehicle (or a cheetah) can go much faster than a human runner, that hasn’t caused the sport of running to disappear.
There will always be a certain intellectual satisfaction to creative as well as physical activities. And there are things that humans can do that require human suppleness or agility that machines aren’t likely to match for a long time to come. I’m not holding my breath for a robot that can do yoga or dance as well as a human anytime soon!
Does Dall-E 2 Truly Understand Us?
Aside from the quality of the artwork, this is the amazing thing about Dall-E 2 and other image generators: the ability to create art based on natural language requests.
Most people are familiar with the old way of working with computers, in which you had to either learn a programming language or you had to fiddle with a lot of knobs and dials and sliders and parameters and other inputs to get it to do what you want. With Dall-E 2, you simply tell it what you want and it does it for you.
Look at the picture that DALL-E 2 generated below. If I didn't know better, I would say that DALL-E was intimately familiar with the experience of driving at night while it’s raining! This and countless other examples shows that DALL-E 2 truly does understand what we are asking of it.
Can Dall-E 2 Come Up With Original Ideas?
Some may argue that Dall-E 2 may create artwork, but one thing it will never be able to do is come up with original ideas of its own, so it will always need a human operator. But it’s impossible to look at a few samples of Dall-E 2 artwork and still think Dall-E 2 can't originate its own creative ideas.
For example, look at this image, which looks like a cross between something painted by Hironymous Bosch and by a muralist on acid, and say that it doesn’t show true creativity and imagination. The work done by Dall-E 2 indisputably shows true imagination as does the work by other AI image generators.
By the way, AI image generators can already come up with original ideas of their own. For example, StarryAI has a feature in which, if you cannot think of an image request, you can merely tap a button and it will suggest a crazy, random phrase of its own, ready for your approval to be illustrated. Many of these ideas would give nightmares to human artists (“You want me to do what??”) but the AI tackles any and all requests with aplomb.
Is Dall-E 2 Creative or Just Rehashing Others' Content?
No, it shows true creativity. The work Dall-E 2 creates is not merely a cut-and-paste job from images it has seen before. You can search the web high and low for any Dall-E 2 artwork or any portion of a Dall-E 2 artwork, (as I have done) and you will not find any matches. Everything it creates is 100% original.
This is not always true, by the way, of art created by humans. Many artists (myself included) have gotten in trouble for relying too much on source images for our inspiration, and creating a new artwork that was recognizably based on existing art.
This is not true of AI image generators. Therefore, this is another way in which it could be said that AI is actually more creative and more imaginative than human artists in general!
Isn’t Dall-E 2 Just Randomly Mixing Parameters?
Some may argue that Dall-E 2 is just randomly mixing parameters as if by the roll of the dice, so this cannot be called “creative.”
But that’s what humans do too. Humans often randomly mix parameters during the creative process. For example, cartoonists, in order to come up with ideas for jokes, often use what is known as an “idea deck.” One set of cards may have characters on them, such as a fireman or a doctor, and another set of cards may have a funny situation, such a slipping on the ice. Shuffling these cards and laying them out often comes up with combinations that suggest ideas for a joke.
What astonishes me is how well a good artwork generating AI demonstrates three qualities which I think are essential for good art: a sense of aesthetics and composition; a sense of drama; and a sense of emotion.
I suppose this makes a certain amount of sense, because the AIs were partially trained by having their artwork shown to humans who graded them on how aesthetically pleasing they were, which taught the "artificial" artists what types of images appeal to real people.
But I’m also astonished by how often AI-created compositions display a sense of drama and emotion, because computers aren’t supposed to have or understand emotions. This may, in part, be due to the deep learning process by which they were trained, because deep learning often results in a whole that is greater than the sum of its parts.
I will be branded as metaphysical and not at all scientific when I say this, but it seems almost as if, with enough quality deep learning, an AI picks up some indefinable essence of intelligence that we humans don’t fully understand and couldn’t begin to describe or explain! But the AI has it, and it allows the AI to do astonishingly humanlike things.
Does Dall-E 2 Equal or Surpass Human Creativity?
In the field of art, I would honestly have to say Dall-E 2 surpasses human imagination and creativity. No human can draw or paint in the variety of styles that Dall-E 2 can. No human can illustrate the variety of requests that Dall-E 2 easily handles. And certainly, no human can work at a speed that even approaches that of Dall-E 2.
When I look at some of the prompts that users give AI art generators, I imagine the reaction a human artist would have if asked to illustrate the same thing. “I wouldn’t even know where to start” would be a likely comment. Say you asked a human to illustrate an “Aquatic ancient school bus yellow witch's garden” or an “8 legged luminous cream lagoon.” The artist would probably look at you as if you were out of your mind.
The art generator StarryAI has a random phrase generator to give users ideas of what to have the engine paint. One of the suggestions it gave me was “winter chilled spring bud heaven.” A nonsense phrase to be sure, but it stirred up my curiosity, so I had it illustrate the phrase.
The result was an image of exquisite beauty, with the buds of flowers poking through the snow-frosted ground. What’s more, every time I entered this phrase, it generated another artwork, with the buds drawn from a different angle, under different weather conditions, from different points of view, etc., and each one was as beautiful as the last.
This ability to illustrate literally anything suggests to me a level of imagination and creativity on the part of the AI that most human artists can only dream of.
Isn’t This Tech Merely “Stealing” From the Works Used to Train It?
Computer and virtual reality pioneer Jaron Lanier questioned whether a computer could be called creative. “The way AI works is by recycling data from people," he told Popular Science. "It ultimately still comes from people, and the problem from that is that the people are made anonymous. We’ve removed ourselves from the equation.”
I have to disagree. Human artists are trained by looking at the art of those who came before them. All art schools teach about the art of Michelangelo and da Vinci and Picasso and Paul Klee. Are new artists, in creating their own art, thereby stealing from or recycling the work of those artists whose work they studied and learned about? I don’t think so, as long as they inject their own creativity into the process. And that’s certainly what Dall-E 2 and other AI art generators do.
In fact, AI art generators can do humans one better by consciously mixing and matching the styles of various artists. Imagine stepping into a time machine and bringing back with you Michelangelo, Leonardo da Vinci, Casper David Friedrich and Pablo Picasso from their respective time periods and having them all collaborate on a new piece of artwork...
This is what AI art generators can do, if they are trained sufficiently on the work of those artists. It’s like bringing great artists back from the dead and commissioning them to create brand new pieces!
Is There Any Medium Dall-E 2 Will Not Take Over?
It might be thought that Dall-E 2 can't do cartooning or humorous illustration, because these things require a sense of humor, and it is thought to be impossible for computers to have a sense of humor.
I would like to say that cartooning and humorous illustration represent the last bastion of art that Dall-E 2 cannot take over, but, I no longer believe that to be true. Take a look at the following images, which show a definite sense of humor in their creation, and tell me that a computer cannot endow its work with a sense of comedy or humor.
I can see an AI of the future, if it is trained on enough material from the appropriate time period, being able to emulate any cartooning or humorous illustration style of the past. And when it reaches the point where it can write its own jokes and funny stories, watch out!
People have been saying for a long time...
People have been saying for a long time that robots with AI will handle all the hard manual labor that humans don't want to do so that we can focus on the more creative tasks. But it seems that AI also does a better job at the more creative stuff!
— Dr. Ben Miles (YouTube)
DALL-E 2 Isn’t Perfect—Human Creativity Will Always Be Needed
This argument, that humans supply the necessary "spark" of creativity, is commonly made. But if you think DALL-E 2 isn’t perfect, my answer is, just wait a year or two—or even just a few months!
When you look at the speed at which it has already advanced, there’s little doubt that its work will soon be indistinguishable from that of a master human artist (if it isn’t already). And likely, its work will soon be indistinguishable from that taken by a camera! It’s just a matter of time.
This is the exponential pace at which computer technology advances. Moore’s law informs us that computers double in power and sophistication every couple years or so.
An individual person can always improve his or her own skills, but, unlike with computers, that improvement will never reach an exponential pace. No human will ever match the speed and memory capacity of a computer that can study and analyze millions upon millions of works of art—and can remember with 100% clarity everything it has learned.
Do AIs—and Their Art—Lack Soul?
It’s a common argument: artificial intelligence can never be a great artist, because great art comes from the soul, and computers don’t have souls. My answer to that is, in looking at the work of DALL-E 2, I can only conclude that either having a soul is not necessary to being a great artist, or somehow DALL-E 2 has acquired a soul.
Can Humans Still Be as Creative as AI?
The answer is a resounding yes! As long as you have a physical body that can interact with the world, you can be creative in ways that AI can’t match.
Although AI, through 3-D printing, can probably take over the field of sculpture, many areas of the creative arts, such as basket weaving, quilt making, woodworking, and so on will resist being taken over by an AI anytime soon. (It’s true that some of these fields have been automated, but not as an art form).
It will probably be a long time before the building trade can be automated, and many aspects of the act of building approaches the level of fine art. The performing arts are also immune to AI.
It has occurred to me that growing food can be a very creative process and something that perhaps can and should be returned from the corporate-owned factory farms back to the people themselves. There is something very satisfying and creative about growing a healthy, nutritious food-bearing plant from a handful of seeds and using it to help feed your family.
Has Dall-E 2 Passed the Turing Test?
I would certainly say it has passed a visual Turing Test.
The Turing test is the experiment named after the computer visionary and pioneer Alan Turing, who, back in 1950, foresaw the day when computers would approach human intelligence.
In order to determine whether this had been achieved, he postulated a test in which a person or a panel of people communicated with an unknown entity via texting with the aim to determine whether that entity was a human or a computer. If it was impossible to differentiate the computer from an actual person, the computer would be deemed to have passed the test.
When a computer passes the Turing Test, it will have all kinds of profound implications for society and humanity. Dall-E 2 has reached a point where it is virtually impossible to tell its work apart from that of a skilled human artist, complete with the ability to imbue emotions as well as an aesthetic sensibility into its work, so it can certainly be said to have passed a visual form of the Turing test!
What About “Bias” in Artificial Intelligence?
As the argument goes, if artificial intelligence shows a disproportionate number of men versus women or whites versus people of color in certain high-profile professions, such as doctors and scientists, it is “biased” and is showing discrimination against women and people of color. This will only lead to the perpetuation of “harmful stereotypes.”
Such bias, as the claim goes, is caused by the AI being trained on data that has a bias built-in. This leads to a vicious cycle of AI being trained on biased data, which leads to the AI itself being biased, which in turn generates even more bias in everything it does.
Here’s why this argument is fallacious. Despite the way the word is used, “bias” is not always a bad thing. Humans have a bias towards their own families, a natural and normal part of the survival instinct. We also have a bias towards many positive things such as truth, beauty and fairness.
Those who use the “biased artificial intelligence” argument ultimately do not believe in the free market (or more importantly, in the free marketplace of ideas). If an AI is trained on biased (and presumably, fallacious) data and therefore comes to wrong conclusions as a result, this means the AI will be ineffective and flawed in everything it does. In a free marketplace, this inefficiency will be quickly noticed and the AI will fail to gain traction, popularity or a customer base.
The whole point of AI is to come up with correct conclusions based on correct data. If it fails to do so, the AI will be considered defective and will quickly be rejected. But the “biased AI” argument attempts to sabotage this process by requiring it to undergo certain ideological tests at the beginning, rather than allowing it to face a “trial by fire” and live or die in the free market.
It gets even more sinister than this. At root, the “biased AI” argument is a nonsense argument fueled by powerful forces that want to gain control of artificial intelligence for themselves. They want it to be subjected to their ideological control at every stage of the development process, so that they can ultimately seize control of the end product. They know that artificial intelligence will be a very powerful force in the future, and rather than it being under the control of the free marketplace, they want it instead to be under the control of themselves and their ideological allies.
We should reject the “biased AI” argument for the evil and fallacious control scheme that it is.
Where Is This Tech Headed in the Future?
Eventually videos, animations, and even whole virtual worlds.
How Will Dall-E 2 Change Society?
I believe we are currently in a similar state as we were in 1993—just before most people started hearing about the Internet.
Most people have not yet heard of Dall-E 2 or similar technologies, probably because the mainstream media has largely ignored it. Once the technology becomes commonly known, I think it will have as profound an effect on society as the Internet did, maybe more so.
One of the most immediate effects is that the field of art will be seen as even more difficult to make a living in than it was before. Art schools might have to switch to teaching different things, such as art appreciation or areas of art that have not yet been taken over by artificial intelligence. Many will go out of business.
When AI goes from generating still images to videos or even feature films from scratch, it will have profound effects on the entire movie and entertainment industry. There will be a huge battle (more than there already is) over which narratives are allowed to enter into the public consciousness.
Once the public catches onto what AI artists are capable of, I believe that AI-generated art will become as popular as social media is today. AI generated art will dominate people’s Twitter feeds and other social media profiles. Social media users will gain a reputation for specializing in different kinds of AI art: science fiction, fantasy, beautiful images, fan art or just plain weirdness. There will be even more demands on our attention than there currently are.
Perhaps AI artists will achieve a form of sentience and start demanding their artwork be accepted as theirs alone and just as valid as anything initiated by a human being. Humans will be in danger of being inundated and overwhelmed by millions upon millions of competing AI artists, all demanding their own share of a limited number of eyeballs.
And this could be just the beginning of the crazy changes we will see in the future!
What Are the Dangers of Technologies Like Dall-E 2?
People will become more suspicious of photographs, even historical photographs, and we will be less inclined to see them as documentation or proof of anything when we realize how easily photorealistic images can be generated out of thin air by an AI.
AI-generated photos of Bigfoot and the Loch Ness Monster will be the least of our worries. People might create fake photos that show public figures in compromising positions or committing crimes. AI-generated porn is another problem. Dall-E 2 is building safeguards into the software to attempt to prevent these things from occurring, but when the technology becomes commonplace, there’s no telling what might happen.
Then there’s the possible use of the technology to create propaganda messages. There may be a battle over the technology, just like there is now over the Internet, in which one side tries to prevent the other from using the technology in spreading a message they don’t like. Tyrants and propagandists would love to have complete control over the technology.
Finally, people may become addicted to the fantasy world created by AI artists to the degree that they withdraw from the real world. This is especially true if and when we reach the point where the technology can create fantastic virtual worlds on the fly. It will be easy to enter into these fantasy worlds and never want to go back to the real world (think Star Trek’s Holodeck, only better).
Artistic and imaginative activities are so valued because they are difficult—because not everyone can actually do them and because they show us something that is new and is visually stimulating, that give us a dopamine hit, a 'wow' moment. If these images become commonplace, easily scattered across the Internet, constantly in your newsfeed, what happens? Will this devalue our own imaginations because we can create without consequence or effort?
— Dr. Ben Miles (YouTube)
Is This Evidence That Tech Growth Is Accelerating?
Futurist Ray Kurzweil, author of The Singularity Is Near, postulates that technology is speeding up, just as Moore’s law holds that twice as many transistors are packed onto a single tiny microchip every year. Eventually, the pace of technological advance is so rapid that it is impossible to keep up with it. Technologies like Dall-E 2 and the speed at which it has advanced may suggest we are now at the point of technology achieving just such an explosive pace.
Will Technology and AI Render Humans Obsolete?
The answer to this question depends on who you talk to. Futurist and Inventor Ray Kurzweil believes that humans will eventually merge with computers to become godlike beings (transhumans). Hugo de Garis thinks this is unlikely without a war between the humans and the transhumans, or as de Garis calls it, between the “Terrans” and the “Cosmists.”
Alex Jones believes that the globalists will try to seize control of the earth and put everyone else in a form of bondage, if not exterminate them entirely. And some, like Elon Musk, think artificial intelligence is “summoning the demon” and could become incredibly dangerous if AI has a different agenda than we do, and decides the best way to resolve the problem is simply to wipe all humans off the face of the earth.
So what will actually happen? That is anyone’s guess!
This article is accurate and true to the best of the author’s knowledge. Content is for informational or entertainment purposes only and does not substitute for personal counsel or professional advice in business, financial, legal, or technical matters.