Midjourney: The AI That Won Over Humans
With the rise of AI-powered tools that can create music, paintings, and even some level of animation, the role of artists is being redefined.
Join the DZone community and get the full member experience.Join For Free
Art-generating artificial intelligence, such as Midjourney, is transforming the world of art. With the rise of AI-powered tools that can create music, paintings, and even some level of animation, the role of artists is being redefined. For centuries, art has been seen as a uniquely human endeavor, requiring creativity, imagination, and emotional intelligence. However, recent advances in machine learning and generative algorithms have given rise to a new generation of artificial intelligence that can mimic many of these human attributes.
As we saw in our article focused on DALLE, AI-generated art is not new, but the sophistication and popularity of these tools have grown exponentially in recent years. From portrait-generating algorithms to computer-generated music, machines are creating works that are increasingly difficult to distinguish from those made by humans. Some experts predict that AI-generated art will soon be indistinguishable from human art and that we will soon see a shift in the way we think about art and creativity.
But what does this technological shift mean for the millions of people who depend on creative industries for their livelihoods? Artists, writers, musicians, and other professionals face the threat of automation, with machines able to take over jobs that used to be done by humans. The fear of job loss is not just limited to the creative industries. As AI continues to advance, it is likely to replace jobs in other fields as well, such as marketing, law, finance, programming, etc.
In addition, the rise of AI-generated art also raises new questions about the nature of creativity and authorship. Who owns the rights to a work of art if it was created by an AI? Should machines be considered artists in their own right? And what does this mean for the future of art?
In this article, we will explore the impact of AI-generated art on jobs, artists, and contests that encompass both. In addition, we will explain how to use one of the most famous and, personally, my favorite AI-generated works of art. Finally, we will answer the question: Who do these images belong to?
Can an AI Win an Art and Photography Contest?
Artist Jason Allen won first prize in the Digital Arts and Photography category at the Colorado State Fair, USA, with his work Theatre D’Opera Spatial. And below, I will show you the proof of it:
So far, everything is fine; no one is alarmed. But what if I told you that the image we are looking at was generated by an AI? Allen would have gone unnoticed if he hadn’t posted a message on the Midjourney Discord server as soon as he learned of his victory.
In the message, he explained how he had used artificial intelligence to create the images he submitted to the contest: “I generate images using Midjourney, make corrections using Photoshop, and use Gigapixel to enlarge.” Arguments that he used sometime later to defend himself from all the criticism he received since, according to him, he had involved the human element in the process.
What is art, what does it mean to be an artist, and how much of a future is left for artists now that AI has burst onto the scene? These are complicated questions to answer since we don’t know how far these tools will go and what they will cost in the future.
In addition, the fairness of the contest was raised, as several Twitter users were able to verify, after contacting them, that the judges had no idea of the origin of the image.
Allen finally decided to defend himself from all the criticism by making the following arguments: “So, are you judging the art or the method by which we create art? What if we approached it from the other extreme: if an artist faced a tremendously difficult and complicated set of constraints to create a piece, like a painting, while hanging upside down and someone is whipping him (this is extreme, I know)? He adds, “Should this artist’s work be evaluated differently than another artist who created the same piece normally? I know how this will end: they’ll just end up creating an ‘artificial intelligence art’ category for things like this.” The truth is that we find some logic in his argument, i.e., is a work created by an artist in extreme conditions better than an artist who has the best tools, colors, etc.? Is the final result or the process in which it was obtained valued?
Allen was also accused of cheating by hiding how he got the artwork, but he defended himself by saying that he submitted the work as “Jason Allen using Midjourney.”
Now, let’s explain the AI that caused all this fuss, how to use it, and how it has evolved to the present day.
What Is Midjourney?
As we saw in the previous article about DALL-E, with a simple description, there are tools that generate art, something revolutionary. This powerful technology is having a great impact on our society, and this is the case of Midjourney, which has not wanted to be less. Midjourney is an independent artificial intelligence research laboratory that has developed a system that it has decided to name after its own company. Like DALL-E, Midjourney is used to create images from a text description (Prompts).
It is true that how Midjourney works from behind is a bit fuzzy, and it is speculated that the underlying technology is based on Stable Diffusion, another AI that we will talk about in one of our next articles.
The style of each of these AIs is unique and particular. For example, the quality of Midjourney is amazing. It creates canvas-like artwork, something more artistic with random human-like details. In contrast, DALL-E 2, in general, gives you a more realistic result.
Another difference is that Midjourney can enhance AI-created images up to a maximum resolution of 2048 x 2048, while DALL-E 2 is stuck at 1024 x 1024.
Midjourney has a free but limited option where you can make about 25 images and then has different plans ranging from $10 to $60 that charge you for GPU usage. Each image generation consumes some of this GPU time.
For the moment, Midjourney is still a beta that was opened to the public in July 2022, so we can only use it as a Discord bot.
How To Use Midjourney
The first thing we will have to do is to create an account in Discord. It does not have any secret. It will ask us for an email, a user, and some personal data. Then we will only have to validate our email, and that’s it. We already have a Discord account.
Now we go to the Midjourney website, and we will find this main page.
As a curiosity, if we click on the eye that we find below in the center, we will see different images that have been developed with Midjourney.
To continue, we will click on the Join the Beta button, which will redirect us to a screen of the Discord web that will say that we have received an invitation to join the Midjourney server. We will accept the invitation, and we will find ourselves on the following screen:
As we can see, we have quite a few options, so let’s go step by step. On the left, we find the Midjourney server and all its text channels, from chats between the different users of the platform to queries that the Midjourney team will answer. To the right of this, we have a small spoiler of the potential of this AI that, as I said, is brutal. I would not know how to differentiate this art from one made by a human being.
We are interested in the section called NEWCOMMER ROOMS 3 (the number may vary depending on the account and when you access it). Within this section, we have a couple of text channels called newbies. This is where we can do our testing.
We will find many users generating their images but don’t worry, the bot will tag us in our result, so we will be notified when we get it. If we go to the text bar and type “/,” we will see a lot of commands. The one we are interested in is called Imagine.
We will type “/imagine” and press the Enter key to get the following:
Inside the prompt is where we can write what we want the AI to draw. We are going to continue with our example of the robot, but this time we are going to tell it to draw it as if it were a villain. It should be noted that Midjourney only understands English, something to keep in mind. I have written the following: “A robot as the villain of a Disney story, digital art”
This has been the result. Midjourney also generates four images, but it also gives you several options. The buttons with the letter U and a number are to tell the bot to return the image scaled to the maximum resolution (upscale). The button with the blue circle icon is for the bot to generate an image again with the same prompt if we are not satisfied with the result. Finally, the buttons consisting of a V and a number are used to generate different versions based on one of the four results. I personally liked the first image, so we will ask it to return to us in high resolution by pressing the first button of all.
This way, we will get the image in a good resolution. Now, for example, let’s try to make some versions of the selected image by clicking on the Make Variations button.
As you can see, he has followed the style and composition of the main image and has created four different versions of it. We could do this process to refine the result and get what we are looking for but to be honest, the first result he has given us is already very good for this simple example we are doing. What name would you give to our villain?
This is a very poor example compared to the full potential of Midjourney and all the possibilities it offers. The last thing they have presented is their new generative model called Midjourney V5, which provides us with very high consistency, excels in the interpretation of natural language prompts, has a higher resolution, and supports advanced features such as repetitive patterns capable of creating borders. Another functionality is another model called Niji, which focuses on producing anime and illustrative styles.
This model is much more knowledgeable about anime styles and aesthetics and is excellent at dynamic action shots and character-centric compositions in general. Finally, something fascinating is the command called Blend, which allows us to load 2 to 5 images, and then Midjourney analyzes the concepts and aesthetics of each image and merges them into a new novel image.
If you want to learn more about the possibilities of this AI and really test Midjourney, it has very well-structured and explained documentation that can be found here.
The Midjourney Community
Our images will be stored in our Midjourney user profile. To see an example, we only have to click on the Sign In button on the main screen of their website and agree to register with our Discord user.
This is our profile, and these are all the images that we have generated with our users through Discord. On the left, we will see a few options since our user has a free version. If we click on Community Showcase, we will see some of the images that other users have taken so far.
Something that the paid version provides you with, which is very interesting, is the ability to access the information on how these images have been generated, such as the prompt used, the configuration used by the user, etc. In addition, several more options are unlocked, such as a search engine with which we could search for “robot” and get all the images that have been generated and contain the word robot.
This could be useful, for example, to get inspiration, learn how to generate better images of what we want to achieve or copy the prompt of that user to try to generate something similar but unique for us.
Art-generating artificial intelligences are transforming the world of art and creativity. While some fear that machines will replace human artists and render certain professions obsolete, others see this technology as an opportunity to explore new creative possibilities and push the boundaries of what is possible.
One group of artists who may be particularly affected by AI-generated art is concept artists. These professionals are dedicated to creating sketches, illustrations, and other visual representations of ideas for movies, video games, and other creative projects. While AI may not replace concept artists entirely, these artists will likely need to adapt to the new technological landscape and find ways to incorporate AI-generated art into their work. Some of them may even choose to use AI themselves to enhance their creative process and generate new ideas.
Other platforms that may be affected are image banks, as it is now perhaps more profitable to have an AI generate the image you are looking for instead of paying for it on a website like Shutterstock. Does this mean that art is going to die? In my humble opinion, no. This debate has come up many times, and, in the end, neither photography nor cinema killed realistic painting.
In addition, AI-generated art has also given rise to a new type of job: prompt engineer. These professionals specialize in understanding AIs and knowing how to get the most out of the prompts these AIs use to generate the content we’ve been talking about in these articles.
One question we often ask ourselves is who owns the rights to a work of art if it was created by an AI? Well, the truth is that currently, no one does. As such, a copyrighted work “must be created by a human being,” and the United States Copyright Office (USCO) says it will not register “machine-generated” works that lack the intervention or creative input of a human author.
In conclusion, the rise of art-generating artificial intelligences is transforming the world of art and creativity in ways that we are just beginning to understand. While there may be some challenges and uncertainties ahead, there is also a wealth of opportunity and potential for those willing to embrace this new technology and push the boundaries of what is possible. As AI continues to advance, we can expect to see new types of art and artists emerge, along with exciting new possibilities for the future of creativity.
Published at DZone with permission of Isaac Alvarez. See the original article here.
Opinions expressed by DZone contributors are their own.
Using Render Log Streams to Log to Papertrail
Database Integration Tests With Spring Boot and Testcontainers
How To Use Geo-Partitioning to Comply With Data Regulations and Deliver Low Latency Globally
Seven Steps To Deploy Kedro Pipelines on Amazon EMR