This article was edited for clarity and minor corrections on 12/20/22
Why generative AI is awesome and awful and what to do about it.
AI image generators like Midjourney, Dall-E, and Stable Diffusion as well as new apps like Lensa are all the rage right now along with other AI generative tools like Chat GPT which can write text of all sorts based on simple prompts. The sudden explosion in popularity and increased quality of these tools has generated a lot of reactions ranging from creatives fearing for their livelihoods, to debates about what constitutes art, and lots of people ecstatic to dive into this new form of creative computing.
I’ve been somewhat fascinated by this tech for a few years and have recently spent several months diving into one platform in particular: Midjourney. I was drawn to it due to its implicit bias for beauty. I was also intrigued by how it uses a Discord server as the front end, and the communal iteration component that creates. The use of this tech is full of both amazing opportunities and complex ethical and legal challenges. I am going to share some of my thoughts on this starting with what it is, what I love about it, its current shortcomings, and what is very concerning. In full disclosure, I am also using AI to help me write early drafts of some portions of this article. If you want to see what those unedited generations were, and the prompts I used to generate them, they are here. Additionally all the images in this blog post were generated by AI in Midjourney via prompts I provided and spent way too much time on.
How it started
Artificial intelligence (AI) technologies that can generate images from text descriptions have been an active area of research and development for decades. The ability to generate images from text descriptions has a wide array of applications, such as creating virtual reality experiences, improving natural language processing, art, and illustration, as well as aiding in the design and visualization of products. The list could go on and on.
The origins of AI text-to-image technologies can be traced back to the early days of computer science and artificial intelligence research. In the 1950s, researchers began exploring the use of computers for image recognition and machine learning. This early work laid the foundation for more advanced techniques that have been developed over the years, such as convolutional neural networks and generative adversarial networks. Feel free to explore the wikis on those if you want to be as confused as I am about what they actually are.
To put it simply, one of the key early developments in AI text-to-image technologies was the use of generative models to create images from text descriptions. Generative models are a type of machine learning algorithm that can learn to generate new data, such as images, based on a set of training data. By learning the patterns and relationships in a dataset, generative models can produce new, synthetic data that is similar to the training data.
In the 1980s and 1990s, researchers began exploring the use of generative models for text-to-image generation. These early efforts focused on generating simple images, such as line drawings and simple shapes, from text descriptions. In the 2000s, researchers began to apply more advanced techniques, such as deep learning and neural networks, to the problem of text-to-image generation, leading to significant improvements in the quality and realism of the generated images. Of course, the quality of results from any such network is also contingent on the data it is trained on. The best source data for visuals is typically going to be from quality image sets. Not surprisingly those are likely to come from professional skilled artists and photographers. This leads to ethical and legal problems around making something that creates awesome results.
Understanding the basics of what is going on here is a key point used in the defense of AI-generated art so, whether or not we agree with that argument, it is worth noting what that process is. Think about how our human brains work as an analogy. As we enter the world we begin to see things and learn to associate language with them. Over time as toddlers, we learn how the abstract word dog corresponds to an actual dog or a picture of a dog. With practice, we can start to draw a dog based on the millions of exposures we have had to various dogs. An individual may decide to pursue the study of art. In the course of their education, they will learn about many artists, their works, and their styles and probably learn to imitate some of them while ultimately working toward honing their craft and developing their own style. Everyone’s “unique” style is in some way a synthesis of all of these exposures combined with their drive to express something. Humans posses a capacity for creativity which can be hard to define. Some artists try to apply their creativity to be as unique and original as possible while others may clearly primarily copy an existing style. Judgment abounds from many corners of society on every aspect of this with humans and with machines even more so it seems.
AI is like that but in some ways, exponentially stronger due to being able to be rapidly trained on millions to billions of works and then rapidly execute a vast amount of variations. In other areas, it is significantly weaker than our human cognition and creativity. It will after all, only do what it is asked to (sort of) do and arguably has no real personal experience, sense of struggle, or soul and no innate drive to express itself. Recent advancements have put this tool into the hands of the masses creating a scale of creative capital never before encountered in known history.
Today, AI text-to-image technologies are starting to be used in a wide range of applications, including virtual reality, natural language processing, and product design. These technologies continue to evolve and improve, and researchers are working to develop even more advanced methods for generating high-quality images from text descriptions. Love it or hate it, AI is here in a big way almost overnight.
I think the interesting thing to consider here is, that we are heading towards a reality where so much of what we consider desk work will be able to be automated. Especially and perhaps in particular digital work. This includes design related things like visual art, illustration, and photography which is what we largely see happening right now with text-to-image generation. It has not quite yet moved to what we would consider able to replace graphic design or branding whole cloth however we must presume many tasks will be affected dramatically and that it’s only a heartbeat away.
For example, we have AI-generated text that can write marketing copy, and it’s not a far leap to foresee AI-generated brand guidelines. I used Chat GPT (which got a million users in a week) to write the first drafts of some parts of this article. Doing so enabled me to learn new relative details to consider and include. I couldn’t just plug that in and hit publish as it lacked my personal style, and needed to be fact-checked and elaborated on. But it quickly proved its value as a tool and collaborator. I’ve also taken AI-generated text and plugged it into AI text-to-image generators as an experiment. This is where I think things will get very interesting. Not just the newbie surface level of dropping a prompt into a generator and using what it spits out as a final design. But using it as a tool to create bits and pieces, concepts and ideas. I think about interlinked AIs working together to perform more complex interdependent operations that can lead to much more complex and targeted results.
AI can be used to generate elements that are then worked into more complex final designs as building blocks. Just as creators may use Photoshop filters, stock photos, clip art, mockups, and the like as part of their process and workflow, and just as those were also demonized on arrival, AI text-to-image is another tool in this arena. But it’s a much more powerful one. Much. More. Powerful.
If stock design tools, photoshop filters, and clip art are screwdrivers, AI is poised to become a whole Dewalt power tool combo set with a really smart robot to operate them. Entering this level of use the human creator collaborator is using a tool of immense power, unlike anything we have ever encountered. A tool that is accessible to the masses at a scale we have never before encountered. And the skills barrier to entry is very low to get started. Having spent months entrenched in it, I can say it’s easy to learn, difficult to master and the rabbit hole goes very deep. That brings with it huge opportunities, creative classism, and big challenges.
How it’s going
For my field at the intersection of strategy, design, experience, and marketing I envision all of these pieces working in concert together with big Data to create profound results that could be utopian or dystopian. Isn’t it always a bit of both? In today’s world where almost everyone’s job title includes “content creator,” we will see this continuing to amplify the quantity and quality of never-ending infotainment.
Generative artificial intelligence (AI) technologies are being applied in a wide range of fields and industries. Some examples of areas where generative AI is being used include:
Virtual reality and augmented reality: to create realistic 3D environments and objects for use in virtual and augmented reality environments. That has commercial implications in industries like real estate and tourism.
Natural language processing: to generate human-like text or speech from a given prompt, which can be useful for tasks such as translation, summarization, and dialogue generation. The recent version of Mid journey now utilized NLP to “better” understand prompts.
Product design: to design and optimize products, such as vehicles or consumer electronics, by generating and testing numerous design variations and selecting the best ones based on certain criteria.
Marketing and advertising: to create personalized content, such as targeted advertisements or customized landing pages, for marketing and advertising purposes.
Art and design: to create original works of art, such as paintings, and video, by learning from and synthesizing existing works.
Music & Video: to produce elements, effects, full compositions, and even whole albums.
Overall, generative AI is being applied in a variety of fields and industries, and its potential applications are constantly expanding as the technology continues to evolve.
So in the field of design, we can use AI to generate nearly infinite visual concepts and then refine those into final designs including brand elements such as illustrations, logos, icons, photos, etc. Already, we are becoming able to also do motion graphics & video, scalable vector graphics, all generated by AI and promising to rival anything made by humans. Just as I was writing this blog, I saw the release of a new text to 3D modeling version.
How it's going to be.
It is quite possible that AI could be the solution to our many global crises or the catalyst that ends us.
In other fields, we can expect this tech is being applied to medicine development, chemistry, space travel, network architecture, traffic engineering, national security, warfare, and the list goes on and on. Virtually every aspect of our world will be touched. There is no question that the computational power of AI is far and above that of the individual or team that operates without its assistance.
I think we can expect the first full-length feature film generated completely by AI in our lifetime. Not just a flat-screen film, but a fully immersive and interactive virtual reality environment/game/movie. This is all a matter of when, not if, presuming our species survives long enough to achieve it. Just for fun, I prompted ChatGPT to pitch a movie for me called THE LAST PLANET ON EARTH. I then used Midjourney to generate movie posters for it and a graphic novel concept. In the span of a day, I have a variety of elements that could be used to make a nice presentation to generate interest in the idea.
As someone who makes graphics both traditionally and with digital tools, I get how this makes my peers very nervous and concerned. We’re still far from it eliminating the need for human creativity, it just gave me a super-powered tool to work with to flush out some ideas. For many applications, it certainly does eliminate the need for a specific human currently earning a living in this way. In a way, it further democratizes design and creativity just as desktop publishing did, just as Canva did and many professional creatives who invested a lot of time honing their craft, don’t like being handed over to those who did not invest the blood sweat, and tears.
More questions than answers.
I’ve heard it said that the quality of our results is based on the quality of our questions. So here are some of the questions that are currently generating in my not-so-artificial human intelligence:
If we presume we are headed into a future where human-generated creative output has no commercial monetary value, what does that mean for the nature of creative work, the individual creative or team, and design’s relationship to commerce and daily life?
For that matter if much of all work will be able to be automated or accomplished by technology and human work is reduced to knowing how to apply this technology and maintain it, then what does that mean about our purpose, daily lives, and the need to even work in exchange for money?
If we create a reality we’re creative output has little monetary exchange value… If all work other than the maintenance of society’s physical and digital systems, and the creative application of technology has reduced commercial value, then this calls into question the already precarious notion of capitalism and our way of living that could arguably be made much better or much worse. As I saw put forth in one of the many heated debates in the comments sections of AI art posts: The problem isn’t AI, it’s capitalism.
By extension, if all knowledge is immediately accessible (smartphones) and creativity is immediately generative (Generative AI) and complex problems can be solved in seconds by complex neural nets, then what is the purpose and value of education? Another system already on the verge of a bubble burst from an outdated economic payoff model. How should our systems of education, trades, craftsmanship, etc. be evolving in the face of this new reality?
Another question. In a post-modern world where there really isn’t anything new under the sun creatively and stylistically, is the idea of harnessing past work to synthesize into new novel works completely without merit? Is it more important to preserve our current models of creative commerce and rights to work, or to reimagine all of that in exchange for harnessing all past work into a bold new creative future? I’m not a fan of false dichotomies so let’s presume there is a balance to be found here.
Is the value of the work in the process or the final deliverable? How are they intrinsically linked?
We know design is very process driven and art is often selling the story of the work as much or more so than the actual work. So now that we have the super tools for generating work, does that mean the value is pushed further up the chain to process and story? What about the creative application of these tools, and strategies? Can we not expect AI will also be able to automate thinking better and faster soon enough? If so, what then? How are we thinking about that now?
We can easily shift between utopian and dystopian outcomes here and I think we can expect plenty of both.
These potential new realities could support notions like Universal Basic Income and a world where people can pursue something for the joy of pursuing it rather than in exchange for Commerce. Because we could eliminate the need for commerce if those who consolidate power would ever allow it and work toward it. On the dark side of that, there would still be the need for maintenance workers which leads to classism and a possible Hunger Games-type society. If we’re not careful we could be paving the way, even more so than now, for a small group of ultra-wealthy to enjoy the benefits as the masses are reduced to the slave class.
Problems and Solutions
Orwellian-style cautionary tales aside, let us shift to some hot topics in our artist, creator, and designer communities.
The use of copyrighted material in the training of artificial intelligence (AI) systems raises several legal and ethical concerns. AI systems often require large amounts of data to be trained and tested, and it is common for developers to use publicly available data sources, including copyrighted material, for this purpose. However, the use of copyrighted material in this way may be considered a violation of copyright law if it is done without permission from the copyright holder. The CEO of Midjourney has admitted to using a hundred million images to train their AI model, without consent.
One of the main legal issues surrounding the use of copyrighted material in AI training is the issue of fair use. Fair use is a legal doctrine that allows for the limited use of copyrighted material without permission for certain purposes, such as criticism, commentary, news reporting, teaching, scholarship, and research. To be considered fair use outside of those, the work must be deemed transformative enough to no longer be recognizable as the protected work. Not a cut a dry determination, very much up to interpretation by a judge and precedent. Whether the use of copyrighted material in AI training qualifies as fair use depends on a number of factors, including the purpose of the use, the nature of the copyrighted work, the amount of the work used, and the impact of the use on the potential market for the work. As is often the case with emerging tech, our current legal frameworks did not anticipate this new paradigm.
In addition to legal issues, the use of copyrighted material in AI training also raises ethical concerns. Some people argue that it is unethical to use copyrighted material in AI training without permission, as it may infringe on the rights of the copyright holder and undermine their ability to control the use of their work. Others argue that the use of copyrighted material in AI training may be ethically acceptable if it is done for a legitimate purpose, such as research or education, and if appropriate credit is given to the copyright holder. Still other’s take a more wild west approach of everything is fair game, we’ll do what we want, and get over over it.
Should it be incumbent on the developers and implementors of this tech to follow existing precedent? In general, it is important for AI developers to be mindful of the legal and ethical issues surrounding the use of copyrighted material in AI training. Developers should seek permission from the copyright holder before using copyrighted material in their AI systems, or ensure that their use of the material falls within the bounds of fair use. In many instances that isn’t happening. Furthermore, works are being generated for commercial use through these networks which could be argued is a clear violation of industry standards and law.
How does this affect the value of pursuing art, illustration, photography, and design as a career? I want to touch on how this is affecting current career creatives. I also want to suggest what to do about it.
There is still value in learning art, craft, and design but arguably it will not be as commercially valuable in the same ways, in the very near future. There is no denying this will have massive impacts on the creative fields in ways that we can not even begin to fathom. Why bother learning to paint or draw if I can just have AI do something way more impressive that would take me years to possibly do, and it can do it in seconds?
Well, when I was learning art in my early days it was impressed upon me that through the process of drawing, and physical hand-eye coordination in concert with observation and contemplation, what I was really learning was how to see more deeply. To put it another way, AI can not give you the benefits of taking time to meditate. Deep creative work is meditative and develops our brains in different ways than are critically different from simply consuming creative output. It makes us more fully human and of a higher quality of being. What will be lost in our consciousness by eliminating these practices? I predict we will find out and like all things cyclical, the traditional crafts of art and design will make a resurgence after the wave of AI tech disrupts everything and cycles through our lives and how we create. It is also worth noting, working with AI is a creative process in itself, albeit a different one.
AI can not produce real tangible immersive experiences, yet. We can presume at some point these technological advances may become indistinguishable from what we consider reality. Considering such futures gives more credence to the assertion that we all live in a simulation. But for today at least, AI can not reproduce the immersive international art and design retreat I am co-organizing in Spain next summer. Though it certainly could just as shamelessly plug it. This points to where value is being pushed.
Now let me speak to the protection of copyright and the inherent infringement that is baked into the system. All of these generative AI models are trained on the previous work of others. Remember early rumblings that by using social media networks to showcase your work, you were granting them rights to use the uploaded image of your work? Perhaps these terms of use agreements are an important legal detail in this context.
From a practical standpoint, the AI has to be fed tagged imagery to be able to understand what a lemur looks like if someone asks for a lemur playing the accordion. From a stylistic standpoint, to get the best results, the best work to train it on would be the best work of the best artists. This often has included copyrighted works, used without permission.
There have been several instances where developers and trainers of artificial intelligence (AI) systems have failed to take proper steps to use copyrighted works in a fair and legal way. One example is the use of copyrighted material in the training of AI systems without permission from the copyright holder. This can occur when developers use publicly available data sources, such as online databases or websites, to train and test their AI systems. Often these data sources contain copyrighted material that has been used without permission. In some cases, the use of copyrighted material in this way may be considered a violation of copyright law, even if the material is being used for a legitimate purpose, such as research or education. All the more so as users as generating new derivative works for commercial use.
AI systems that use machine learning techniques to generate new text or images based on existing copyrighted works may be considered a violation of copyright law if they are used without permission from the copyright holder. In these cases, it is important for developers and trainers to obtain permission from the copyright holder or ensure that their use of the material falls within the bounds of fair use. This isn’t happening.
Overall, it should be important for AI developers and trainers to be mindful of the legal and ethical issues surrounding the use of copyrighted material in their work. They should seek permission from the copyright holder before using copyrighted material in their AI systems or ensure that their use of the material falls within the bounds of fair use. Since they are largely not doing this, then the question remains what to do about it?
This also leads to levels of infringement that are hard to decipher but I think we can expect a reckoning to occur here, as it should. There are some similar occurrences in the past we can look to and also some key differences. Ultimately I do not believe proselytizing on ethics will sufficiently address the issues to affect fair resolutions here. Too many people simply don’t care. There are two things that will more adequately address this: legislation and litigation.
Music as inspiration
In the 1980s sampling hit the music scene. Suddenly we had consumer-level access to technology that allowed sampling a piece of existing music and assigning it as instrumentation in the creation of new music. It’s also worth noting that you can’t copyright a style of music. Musicians are often inspired by the work of their peers and predecessors and that collective creative consciousness propels the whole craft forward. We can draw a similar relationship between referencing styles and referencing specific artists’ specific works of art. We see the artist Weird Al Yankovich has made a career out of copying other musicians’ songs while changing the lyrics in a cheeky satirical way. Now you know you’ve “made it” if Weird Al takes on one of your songs. It is worth noting that he negotiates fair use of copyright with the original artists to do so and he would certainly be toppled by lawsuits if he did not.
We saw that early on artists started sampling the music of other artists and then came those inevitable lawsuits. Those lawsuits will and should happen in this space as well. I’ll get into this more as we go here.
To continue the music industry parallel, associations exist to collect and distribute royalties to member artists. If a musician covers another artist’s song on a recording for sale, there is a process by which to obtain the rights to do so.
ASCAP and BMI manage rights and collect royalties from public institutions, radio, and others to pay out to their member artists. The Harry Fox Agency exists to negotiate rights to cover another artist’s or publisher’s song.
Currently these AI models are putting “sampling” technology into the hands of the masses to “remix” into new works. In some cases, these new works are being used as commercial work. Admittedly, this metaphor is a gross oversimplification of what is really going on with the tech so let me tease that apart a little more. By default, it’s a much more diffused approach. I can prompt for an image of cassette tape woodblock print and it will generate an image based on its understanding of thousands of images of cassettes and tapes and woodblock style prints. One could argue there’s not much afoul there even if those thousands of images are copyrighted works. It’s theoretically displaying its interpretation of averages of all those samples while giving more or less weight to certain variables as directed (and in ways completely unknown). The argument in defense of this is, that’s exactly what human creators do too. On the surface, this doesn’t seem that different from how every artist is influenced by all of the work they have seen and studied. To me the most interesting AI results come from prompting for the unexpected and impossible.
But if I prompt for an image of a cassette tape in the style of Van Gough, now we are narrowing into a specific artist’s style and more dubious ethical waters. But you could also say this is akin to writing a song in the style of another musician. That may not get you much respect amongst other musicians, but it is not something that is legally protected or even necessarily looked down on. In fact, creatives continue to inspire each other in this way and push a style into a collective movement, and as such genres are created and evolve.
We can go a step further and prompt for an image of a cassette tape in the style of the “Starry Night” painting by Van Gough. Now we are really zeroing in on a specific work and this is where I would say a line gets crossed. Now we are covering the song and certainly, there is a strong argument for paying royalties if this generation is used commercially.
Interestingly in prompting Midjourney for the last cassette tape image here, I referenced Van Gough and Salvador Dali. Yet it gave me one with the Great Wave off Kanagawa image by Hokusai. This points to the more complex nature of what the AI is really doing. Since these iconic works are often depicted in proximity to depictions of other iconic works in various online publications, it cross polinated into the output. How does one begin track and credit such a thing?
Now just like we can open photoshop and take a picture of Obama and posterize it for fun that’s fair game. But when you take that and profit from it… well, just ask Shepard Farey how that worked out with the Associated Press suing him for using a source photo to make his iconic HOPE poster. The problem is they owned the rights to that source photo. I think Shepard Farey is fantastic. I actually love his rebellious and law-breaking approach to street art and design. You should absolutely watch the Hulu documentary about him. But my admiration for him was not enough to save him from the legal consequences of his actions. (Sorry, Shepard!)
In that case, the process by which Shephard Farey created that image was controlled by him. Here we have a tech that is built upon a process that is a bit of a magical black box. So who is ethically responsible and who is legally liable: the end-user or the AI company?
Now we get to what I think should be done about it. This is where I emphasize I’m just a guy who loves art and design. I love this tech too, but I think it’s due for an ethical regulatory intervention that respects original artists’ rights. I’ll also say I am not even remotely qualified to give any legal advice. Having said that, here are my suggestions.
One is to establish an entity to negotiate fair use of copyrighted works in artificial intelligence (AI.) Similar to the way music royalties are collected and distributed in the music industry, we could create a licensing agency or collective rights organization. Such an agency or organization could be responsible for negotiating and enforcing licensing agreements with AI developers and trainers as well as end users who wish to use copyrighted material in their work.
The agency or organization could work with copyright holders to establish fair and reasonable licensing terms for the use of their copyrighted material in AI training and development. These terms could include the type of use that is permitted (e.g., research, education, commercial), the duration of the license, and the amount of the copyright holder’s work that can be used. The agency or organization could also collect licensing fees from AI developers and trainers and distribute them to the copyright holders based on the terms of the licensing agreements.
In addition to negotiating and enforcing licensing agreements, the agency or organization could also provide guidance and resources to AI developers and trainers on the legal and ethical issues surrounding the use of copyrighted material in their work. This could include information on fair use principles and best practices for obtaining permission from copyright holders.
Overall, establishing a licensing agency or collective rights organization could help to ensure that AI developers and trainers have a clear and consistent way to obtain permission to use copyrighted material in their work, and that copyright holders are fairly compensated for the use of their work.
Follow the money
Litigation and legislation are likely to play a role in establishing an entity to negotiate fair use of copyrighted works in artificial intelligence (AI) and in addressing the current infringement of artists’ rights.
Lawyers are always going to go for the big bank accounts so the individual user is not a great target unless they are making big bank off AI art. Not to mention we are dealing with a scale of use that you could never enforce against individually. So clearly it’s the AI model companies themselves that need to be taken to task.
Litigation could be used to enforce the rights of copyright holders and to seek damages for any unauthorized use of their works in AI training and development. This could include lawsuits against AI developers and trainers who use copyrighted material without permission, or against companies that profit from the sale of AI systems that use copyrighted material without permission. Litigation could also be used to challenge the legality of certain uses of copyrighted material in AI, such as the use of machine learning techniques to generate new content based on existing copyrighted works.
Legislation could also be used to address the issue of copyright infringement in AI. For example, lawmakers could pass laws that clarify the legal status of AI-generated content, such as whether it is considered a “derivative work” under copyright law. They could also pass laws that establish licensing frameworks for the use of copyrighted material in AI, similar to the way music royalties are collected and distributed in the music industry.
Overall, litigation and legislation will likely play a role in establishing an entity to negotiate fair use of copyrighted works in AI, and in correcting the current infringement of artists’ rights. These legal tools can be used to enforce the rights of copyright holders and to establish clear and consistent rules for the use of copyrighted material in AI.
There are several existing professional organizations that could potentially work together to spearhead initiatives related to the fair use of copyrighted works in artificial intelligence (AI) and the protection of artists’ rights.
One such organization is the Association of Computing Machinery (ACM), which is a professional society for computer scientists and researchers. The ACM could work with its members and other stakeholders to develop guidelines and best practices for the ethical and legal use of copyrighted material in AI training and development. The ACM could also advocate for the rights of copyright holders and work to establish licensing frameworks and other legal protections for their works in the context of AI.
Another organization that could potentially play a role in these initiatives is the Electronic Frontier Foundation (EFF), which is a nonprofit organization that focuses on protecting civil liberties in the digital world. The EFF could work with artists and other stakeholders to advocate for the rights of copyright holders and to develop policies and practices that ensure that their works are fairly compensated for use in AI.
Other organizations that could potentially be involved in these initiatives include professional associations for artists, illustrators, designers, photographers, and other creators, such as the Graphic Artists Guild, AIGA The Professional Association for Design, the Society of Illustrators, and the Writers Guild of America, The International Association of Photographic Art as well as legal and policy organizations, such as the American Bar Association and the Center for Democracy and Technology. By working together, these organizations could help to establish clear and consistent rules for the use of copyrighted material in AI and to protect the rights of artists and creators. The results of such a collaboration would need to be negotiated by the community and legal proceedings but here are some possible conversation starters.
The AI models should be required to be transparent on their source data and pay royalties to every copyrighted source used for every image generated. Similar to how ASCAP and BMI work in the music industry. If we follow that model perhaps creatives would join the association that is the intermediary to get paid.
Artists should be able to opt out of being prompted and/or get larger, more direct royalties when they are prompted by name, even more, when my name and specific work. Levels of usage rights should be clearly communicated in the generation process so users know they are violating the law if they do not comply. Additional licensing is required for commercial use of directly referenced artist prompts.
Restitution in the form of back payments for current and previous infringements. This would likely be administered through class action lawsuits as well as individual lawsuits.
These are just a few ideas and I’m sure there are a ton of others to consider. I would love to hear them.
A hypocrites conclusion
I know, I know. I hear you. I’ve been paying Midjourney to play in their sandbox of art thieves. And I’ve gotten my fair share of side-eye criticism from some of my peers for it as I post weird creations on social media. I want to understand it and to do so I need to play with it. To me, that inherent weirdness and exploring what it can and can’t do is part of what makes it so wonderful. It is very easy to get it to surprise you and very hard to get it to obey you. It’s also pretty addictive.
I even used it for my holiday card this year. As I mentioned before the tech is amazing and so fun to play with and I believe there are ethical and non-ethical ways of engaging with the tech. I have used it to create textures, and sample photos to use in mock-ups as well as to collaborate on generating concepts. Using images to prompt along with text and feeding that back in through multiple iterations leads to really interesting results. It’s not the tech that’s the problem, it’s the people and policies. Just as many wrung their hands that photography would be the death of painting, that Squarespace would be the death of web design, Fiver, Design Pickle, etc. All these things disrupt our industries for sure. But we are creatives, we adapt and envision new ways of harnessing the very things we fear may destroy our way of life. What we fear today revolutionizes how we work, then becomes ordinary tomorrow. We ignore it at our peril.
But for now, I have canceled my Midjourney subscription. I do feel wrong paying into something that I know is being misused willingly by the founders. But I couldn’t resist creating a few more gems just for fun, and for this article.
In conclusion, in the words of Stan Lee “With great power comes great responsibility.” The use of artificial intelligence (AI) in the design industry has the potential to revolutionize the way that design is created and executed in ways we can only begin to imagine. By automating certain tasks, generating novel concepts, and providing support for decision-making, AI can help designers to work more efficiently and produce creative designs at new levels.
However, it is important to consider the ethical implications of using AI in design, including issues related to data privacy, usage rights, and bias, and to ensure that the technology is used in a responsible and transparent manner. It is also crucial to continue researching and developing AI in design in order to maximize its potential benefits in an ethical manner and minimize any potential negative impacts on the industry and society at large. It’s up to us to engage in the steering of this new tech in ways that benefit our society and professions.
To simply claim AI is bad for artists grossly overlooks the potential benefits. To simply claim the ethics don’t matter and that everything is fair game, greatly overlooks the inherent responsibilities. To shame those who are “less talented” that now have a new way to feel creative is just not nice. To make blanket rulings to not allow AI assisted art in competitions and copyrights is an understandable reaction in this moment. Yet it lacks the nuance that AI can play in a larger creative process that still may rely heavily on human craft incorporating this new tech. Perhaps those positions may evolve over time if we can get the ethics and legal issues resolved.
Artists standing on the sidelines shaking fists won’t accomplish much in their favor. Those who don’t care about the ethics, will plow merrily along harnessing the advantages it offers, leaving everyone else behind. Yes it devalues art via supply vs. demand economics and that hurts many creatives right now. But I wonder what lies on the other side of that fear and pain? This new technological revolution is here and the sooner we creatives and the organizations that represent us, step up to help steer its development in a nuanced way, advocating for and accepting its ethical use, the better.