“Barry made you this wonderful story,” my wife said as she handed out the home-printed book I made for my niece on the 4th night of Chanukah.
“Well,” I added, “whether it is wonderful is still in question.”
To which my son wryly interjected, “as is whether you made it.”
First, some context: During the holidays this year, I made a number of personalized, illustrated stories for the children in our family. I asked each in advance: “If you were the hero in your own story, what would be the: setting, challenge, genre”. With that information, I went to ChatGPT, and entered a text prompt like “Write me a short fantasy story with a castle about a boy facing a dragon and how they became friends”.
Immediately ChatGPT gave me a dozen-or-so paragraphs composing a coherent short story. I ran the same prompt a few times until I found a story I liked, then incorporated elements from the rejected stories that stood out. I then laid out the text in a design program so I could print them out as a holiday present.
But first, for each page, I grabbed a sentence and put it in as a text prompt for Midjourney, which is like ChatGPT but for images. For example, “He let the friendly dragon try some of the delicious food from the castle’s kitchen” produced the following illustration:
Could I quickly (about 90 minutes per book) and easily produce personalized stories to delight my extended family? For sure.
Was my son right to challenge their authorship? Absolutely.
But would any of these gifts have existed without my efforts? Not for a second.
So what’s the right way to understand my role in the production of this material and what is the new ethical landscape I need to navigate in order to do so in an ethical way?
MidJourney AI Art Station is a Facebook group (currently with 21,000 members) created for people to “discuss all things A.I. art.” One poster recently shared the following insightful observation in a post entitled “ETHICS & AI ART”.
- “This is art”: To this writer, there was no denying that what we see through this tools is a product that produced an emotion filled with awe, which met their definition of art.
- “We are not artists”: That is, just by using these A.I.-powered creativity tools does not make one an artist. This is where the title for this piece comes from, as they wrote “typing words into a computer, getting art back and calling yourself an artist is like ordering food and calling yourself a chef.”
Before I say more I need to set some additional context. If this conversation does not upset you, if you have no skin in the game, you might not realize how much the sudden emergence of these popular tools is threatening the livelihood of visual artists, writers, journalist and more. I have so far not named the writer of the above post because they do NOT want to be identified outside that Facebook group. Why? “Emotions are high right now,” they wrote, “and a lot of artists are getting blacklisted for even KINDA saying AI art is cool / has a place / isn’t going anywhere.” Another poster responded in this thread: “I’ve been doxxed, have a collection of about 50 death threats and my doctors office was called…”
So as I dive into these ethical waters I recognize, not being a trained visual artist whose career is threatened, I want to tread lightly and with great respect.
I love the metaphor about ordering food not making you a chef. It’s pithy and highlights a variety of concerns: about claiming the labor of others as one’s own, about the sense that someone is unjustly pretending to have skills they don’t have; and something that seems unfair about getting something of such value through such little effort.
However, while I think this raises these topics for discussion, they are the not the final word. More broadly, I think this notion is a false one.
To order food, I pick up the phone and describe what I want – or use an app to make some selections and hit enter. 30-60 minutes later food has arrived. Voila!
Meanwhile, while I described above a quick and dirty way to make personalized, children’s books, they were far from a quality worthy to sell on Etsy. A better example is the original “photo” I needed for a deck for a client about a museum concept last week, which took an hour to produce after working through over a hundred options. Or the cards I am producing for a game about A.I.-art, with some cards requiring work of up to two hours to produce. When I think about just the time spent alone, the metaphor of ordering food just does not hold up.
Unlike with ordering food, when I am working with Midjourney I am developing technical skills in using a piece of new technology, like learning how to use Instagram to build a community around a new book, or using mesh routers to spread strong internet access around my home. I am using creative skills identifying what is and is not working, and like solving a puzzle coming up with new combinations of descriptive words and technical terms unique to the tool in order to bend it to my will. It also requires a design practice based on learning through iteration. These are all real skills and literacies.
Perhaps what is missing here is that A.I.-generated text and art is not like Instagram or a router system. Every new tech likes to claim it is disruptive; every moral panic likes to argue that the latest concern is more dangerous than the last. Perhaps A.I. is different, is actually that much more disruptive and its dangers that unique. Why? Because it is operating at a new level of scale, and as such requires a new framework to understand it.
When I started working on the Web in 1995 – that is, literally building the World Wide Web through hand-typed HTML code one page at a time – I looked for some guidance. I found it in Marshall McLuhan.
McLuhan, who is most famous for saying “the medium is the message,” helped us to understand that, for example, the content produced for television is not what shapes us as much as the very act of watching television (and the type of relationships with content the medium of television enables and constrains). So, for the Web, the actual information found on web sites is less important than how humanity is shaped by the ability to click endlessly across a seemingly-infinite number of globally linked set of information. At least, that the idea.
Another concept from McLuhan that had greater impact of me is about the relationship between human’s and technology. Traditional science fiction often used technology as an allegorical warning about humanity’s disconnection with nature, e.g. Frankenstein, an evil computer, etc. McLuhan, however, saw things different. Technology is not in opposition to the natural order of the world but, rather, an extension. A pencil lets you take your ability to speak and extend its ability to last over time. A telegraph lets your voice carry over a greater distance. In this vein, any technology can be reframed as extension of our human powers.
This idea seeped into cyberpunk and from that to the cultural at large, the idea that technology does not destroy our humanity but instead enhances it and that, over time, it will enhance us to such a point we become post-human, passing through a singularity in time from which there will be no going back.
From this perspective, today’s AI-powered creative tools just let us “do art” in a more powerful and effective way.
The tools make us better artists.
But my son was right. Whether I made those children’s stories is still in question, whether I can claim the role of artist who brought them into existence. Just because they would not have existed without me, it doesn’t mean I can claim them as my own.
So what can I claim?
I go back to Clive Thompson’s now classic analysis of how chess grandmaster Garry Kasparov responded in the late 1990s to losing games to IBM’s new Deep Blue. Many threw up their hands, as if the game was over: our future A.I. overlords have arrived. They had not. But Kasparov did not simply reject the power of computers. He created what he called advanced chess, in which players are assisted by software. The players gets feedback from the computer but ultimately make their own decision.
Kasparov then held competitions in which players can choose whether or not to use computers. The winners were those who not only worked with computers, but were the most effective in HOW they worked with them. In the end he concluded “The most brilliant entities on the planet, in other words (at least when it comes to chess), are neither high-end machines nor high-end humans. They’re average-brained people who are really good at blending their smarts with machine smarts.”
In other words, rather than think about technology as something our bodies incorporate through assimilation – like the Borg – perhaps A.I. is developing – has developed – to the point where we can no longer assimilate them, no longer treat them as resources to colonize, but rather threat them as collaborative partners. Earlier this year, Google engineer Blake Lemoine was fired after making claims that their new A.I. was sentient. I think he was wrong, but on to something still worth noting. Our current state of A.I. might not be sentient, but it could still be a technology at a new scale requiring us to treat it differently.
So rather than assimilate it all McLuhan, perhaps its better to think of A.I. as a separate entity with which we collaborate. And in fact, from an interactive design perspective, what is it exactly we are working on with A.I., on what exactly are we collaborating? The answer: making sense of large sources of data. When you search on Google you are using their A.I. to sort through all pages on the web to find what your text prompt requested. It does not produce something new, just gives you access to something already existing, but something you could not have made sense of on your own.
But what if a Google search did give you something new, something it interpreted through your guidance? What would that be called? What could I term my role in the process?
That is what ChatGPT and Midjourney and tools like that are doing: providing us with opportunities to make sense or and make something with our new world of ginormous data sets. So if we can’t use the metaphor of ordering food to make sense of this, and we can’t use McLuhan’s enhancing humanity model, what can we use?
Is there a model can we draw from that describes one’s identity as being part of a collective that makes art, in which no part can make this type of art without the contributions of the other? What terms can we use for the uniquely powerful role humans play within this collaboration among humans and computers?
A member of the Facebook group MidJourney AI Art Station responded to the Ethics post. “I call myself an AI artist with confidence,” she said. “The AI and I are a team.”
That original Ethics post concluded by highlighting an ethical point:
- “We shouldn’t use living artists in prompts”: When making art in tools like Midjourney, it is common to include text like “in the style of Jeff Koons”. This creates art LIKE that artist. It is also fun, and powerful, to combine styles, like “in the style of Jeff Koons, Studio Ghibili, and Ansel Adam,” to generate something wild unexpected and new. “We shouldn’t be using living artists as prompts,” the author of the post wrote, “especially if we mean to profit from the use of these images.” Pulling from and being inspired by past artists, especially those who are deceased, they see as fair game. That’s what artists have always done. But if they are living, working artists “at some point this is going to affect their income” and should be respect.
Until I read that I have been using living artists in my prompts in the design of my card game. No more. And I’ll start reducing that practice now and be more intentional when (or if) I still do.
I expect humanity is just in the beginning of developing more sophisticated ethics around human collaboration with A.I.s. Other aspects have already been well established – uprooting inherent racial and gender bias, exposing black box reasoning, and more. I have so much more to learn.
But for now, personally, I look forward to learning more about my A.I. collaborators and exploring what wonders we can create, together.
ADDENDUM ONE
There are two pieces I want to call out that also informed my thinking above:
Ian Bogost in the Atlantic argues that these new A.I. tools are not epistemological in nature but aesthetic. That is, they are not tools for accessing the truth but for creativity. I think he is largely right – they are largely tools for creating new things, and playing with the possibilities of machine generated content, and less useful for producing knowledge. But this will shift over time. Google is terrified, and rightfully so.
ChatGPT Is Dumber Than You Think: Treat it like a toy, not a tool
A second piece is this one from the New York Times: How to Use ChatGPT and Still Be a Good Person. It’s subtitle sums it up: “It’s a turning point for artificial intelligence, and we need to take advantage of these tools without causing harm to ourselves or others.”
ADDENDUM TWO
For those new to this world, I thought it would be helpful to share some more visual examples of what I am talking about above, a selection of ways I used Midjourney in the past four weeks. Without examples, it might be hard to imagine how powerful these tools have become.
During the past month I generated over 4,000 images. Why?
1. For my end-of-year business email, I used Midjourney to create a digital holiday card. I used the prompt “an experience designer showers friends with books in an enthusiastic display of affection for the holidays” and eventually that led me to this image:
Note that EVERYONE is white, and that it is centered around a male – neither of which I requested. This highlights the bias in the system. To get a better image I turned “showers friends” to “showers multiracial friends” to get something better:
2. I shared above about the personalized books for family members. Here’s an example of one of the pages.
3. I am writing a new book on Stephen Sondheim and games. I needed a placeholder to use for social media so I under Midjourney to give me some ideas.
4. Sometimes an animated gif or a meme is just not enough! My friends were joking about early Judd Hirsh movies, so I created this fake still from his ninja action film.
5. I was working on a report for a planned museum. We needed to create some specific archival images. Here is one page from the deck. Only two of these were real images – the rest straight from Midjourney. Can you guess which ones are real?
6. I am creating a card game about the rise of A.I.-assisted art production. I call it Uncannny Valley, “a game to welcome our future A.I. overlords”. It’s a game about the new visual literacy required to make sense of A.I.-assisted art. The vast majority of the 4,000+ plus images I generated in the past month were in development of this game. Some of the outtakes I posted on Instagram, which you can see here. Below are a few examples:
