Using Midjourney to explore parallel universes
Using Midjourney to explore parallel universes
Imagining Jimi Hendrix's career had he lived, and the ethical considerations of wish fulfillment in AI.
Last night I decided to do a deep dive into Midjourney to try my hand at AI generated art. My goal was to experiment with emulating photography, and generating scenes from parallel universes. I’m a huge Jimi Hendrix fan, so I decided to envision what his life might have looked like had he not died so young. I started prompting, and Midjourney delivered. I got a pretty significant response to this work on Reddit that got me thinking about the ethics of this which I will explore below, but to start, let’s take a glimpse at what could have been.
Imagining Jimi Hendrix's career had he lived
I’m still astounded by what I was able to produce in such a short period of time. Working on this project was a genuinely moving experience, as AI literally peeled back the veil on a parallel universe where one of my favorite artists was still with us. However this joy was not unbalanced by my own lingering thoughts on what this god like tool means for us mere humans, still living in this dimension.
Thoughts on the process
I have been hesitant to dive in as a practitioner of AI until this moment. As an artist I am both troubled by AI’s growing ability to replace me, and excited by the potential to harness AI to bring to life the visions in my head with lightning speed. It is a literal double edged sword, offering increased creative potential, at a devastating price.
On being replaced
I spent a lot of money to go to college and learn the craft of illustration. I would spend weeks on a single illustration, drafting thumbnails, drawing in pencil, then pen, scanning it into the computer, cleaning up my line work, vectorizing it, and coloring it in digitally. You can see some of my work here. It was an exhaustive process that a program like Midjourney can now replicate in seconds. While I’ll admit I do much less illustration now that I have moved into Product Design, it still makes me question how much time and effort I should be investing in “manual art”, particularly where the final product is consumed digitally.
The ethics of training models
The next question I must face, is if programs like Midjourney are even ethical to use. It was trained on real art, made by real artists who are now having to actively compete in the marketplace of attention and work opportunities, against robots trained to emulate them. AI is narrowing the window of artistic opportunity available for humans, and undercutting them on cost and effort. I’m not sure there’s an ethical approach to using the current suite of technology, such as ChatGPT and Midjourney, that were trained on broad data sets culled from the internet.
I can envision a future world where artists build and maintain their own AI platforms, trained on their own creative output, curating an individual style that only they can produce and sell. One example could see a prolific artist such as the Grateful Dead using their extensive catalogue of live and studio music to train an artificial version of themselves, allowing their estate to produce new concerts into perpetuity. It’s still a dystopian vision, where established artists continue to dominate the marketplace long after their deaths, but a more ethical solution than the AI version of Napster, where everyone is able to replicate and distribute their favorite artists at will. You would need to both ethically source the training model, and control access for who is allowed to produce new content off that model. If Matt Groening fed all 743 Simpsons episode into an AI machine learning model, that then made a weekly episode for the next 100 years, would that be unethical?
Experiments in Midjourney
Despite my reservations, I wanted to jump into experimenting with AI generated art. While I acknowledge I can’t solve the training model issue now, I do think there is a way to approach creating “more ethical” content using AI. My approach with this project was to only generate content that could not be created by an artist in a non artificially intelligent medium. So right off the bat we take away pretty much every form of illustration, or anything that cops an established artistic style. Luckily, Midjourney is getting pretty good at mimicking photography, so I decided to start there.
Now I have a lot of photographer friends, so I offer this disclaimer. I am not diminishing the artistic value or identity of a photographer. Someone using AI to generate photos in the style of Ansel Adams would be violating this thought experiment. I am trying to think of photography in a more editorial style, where the “voice” of the artificial photographer is diminished in favor of showing the subject.
Once I landed on a “medium”, I just needed to identify a subject matter that could not physically be replicated by photographers without AI. That’s how I landed on the notion of ‘what could have been’, and exploring parallel universes.
Parallel Universes
It was merely a decision of fandom to select Jimi Hendrix. I wanted to know what he would have done, and where he would have gone had he not died on September 18, 1970. It was easy to come up with a series of prompts that placed him in the zeitgeist of different eras, while letting Midjourney age him appropriately. One example was:
Imagine a photo of 72 year old Jimi Hendrix winning a grammy in 2012
That’s all it took, and the power of AI took over. Now in some of these cases, I did 1-2 layers of variation where you essentially tell Midjourney to riff on a selected image. In this case I wanted a picture where his eyes were a bit more energized.
The experience of using Midjourney to generate content is much more akin to being a consumer than producer. It’s pure wish fulfillment, delivered almost instantaneously. One can easily imagine a not too distant world, where I could ask a similar product to watch a Jimi Hendrix concert from 1985. I could select the venue and backing band he played with, request certain songs be played, or for a different camera angle, or a change of wardrobe. I fully believe AI will eventually allow us to spin up snippets of parallel universes, as we act as directors of our own entertainment. It’s just a matter of processing power and fidelity at this point.
On the notion of likeness
One of the biggest issues I want to address in this project is the ethical considerations of using someones likeness to generate AI content. This particular project is so moving because it uses an established identity, and its fidelity is so amazing because of the vast training data that Jimi Hendrix and the photographers who captured him from the front row created over fifty years ago. For the record I don’t plan on monetizing this or using Jimi Hendrix’s image as a platform for anything more than sharing my own thoughts. This was wish fulfillment, and I would place myself firmly as the consumer rather than the artist in this specific project. I also deliberately made a choice to only focus on positive images representing this subject. I contemplated producing a “mugshot” image, that explored a downturn in this parallel Jimi’s life. Had he lived he surely would have experienced the same rise and fall that all celebrities of any lengthy tenure face. However, the artificial disparagement of an identity, wether living or dead, is where we firmly cross the rubicon into weaponizing AI in a dangerous and unethical way. Even as a work of fiction, it felt unwise and unnecessary to explore that aspect of this parallel dimension.
In conclusion
I’m unsure of where to go from here. I will probably use this tool again to peel beyond the veil into more universes. It’s hard not to utilize such a miraculous power once discovered. I have a lot more thinking to do on AI and how to best utilize it as an artist. I’d love to hear from you and be challenged on my thinking so far. I hope to share more writings on technology as a whole, and how it might be ethically applied, so please subscribe if you want to follow along. Thank you for reading!