Whether you’re a latent, aspiring, or professional artist (a term I use in the broad sense to encapsulate actors, writers, musicians, sculptors, etc), you may be filled with existential dread as you peer over the precipice of AI-generated and assisted creativity services.
Perhaps you've been spiralling, wondering what will AI do to the future of art? Will AI take over art? Can AI create true art?
You might take a little comfort in knowing that, for the moment, this precipice overlooks an uncanny valley of ersatz AI music sung by robotic vocalists - autotuned to a crisp, landscapes that defy physics, lighting, and perspective, and shiny thirst trap images of women with a dead-eyed gaze and mangled fingers.
Additionally, AI creativity tools also suffer from (surmountable) challenges for their human commissioners; The race to capitalize on AI learning models makes it challenging to compare tools without mining YouTube tutorials or moving past a paywall to see if a given model (or set of AI models) will create the results you’re looking for. This leads to some interesting experimentation, but it also leads to disappointment.
The current challenges of widespread adoption of AI creativity tools include:
- The inability to know if you’ll get the results you’re after without going beyond a paywall (theres a vast difference in output quality between many free and paid versions and some tools don’t provide free trials)
-
The inability for the tool to refine its initial output
- If you request changes to an output, you get an entirely new output rather than a revision of the original
- Some AI tools are starting to offer fix this (for an additional price)
I would liken the experience to paying a freelancer to merely try to create a logo, and if they don’t succeed in ultimately creating what you want, they still get paid.
They supply you with a first pass at a few logo options based on information and preferences you supply in advance. After you weed out options that don’t suit your brand, you provide feedback on one of the logos that comes fairly close to your vision, but it needs some work.
However, unlike when you commission a logo from a freelancer, the AI generates an entirely new logo, incorporating some of your initial feedback, and adding your additional feedback on the logo you mostly liked.
At this point, depending on your payment arrangement, you may have to pay up again and again for the AI to keep trying. Worse yet, they may start to forget some of your feedback. The output revisions get worse over time or you just have to pony up a few more dollars get improvements.
Much of this is due to the fact that the business model for AI creative tools today is mostly token-based rather than outcome-based. This will necessarily shift as processing gets cheaper, models improve, and market consolidation weeds out the wheat from the chaff of competing AI tools.
Additional challenges that are sure to smooth out over time include:
- Society's resistance to change and their slow rate of adoption
- Fear of the future/distrust of AI
- Learning curves around new terms, current options, and prompting and project token optimization
- Poor first impressions from past/current AI capabilities
- Barriers to entry such as techy set ups and poor user UX design of nascent and open-source tools
- The need to cobble multiple AI generators rather than an all-in-one solution
But these challenges can’t hold back the tide of change for long, so you should probably give up now, right?
Wrong. Here are several ways you can succeed as a human artist in a post-human age of artificial-intelligence-generated creativity, which I plan to delve into in more detail in other posts.
-
Use AI as a collaborator - prompt an LLM to give you prompts to overcome writers block or blank canvas syndrome.
When preparing to interview artists for my book, I asked Claude to give me some inspiration for questions I could ask, based on the themes of my book and the background of the interviewee.
-
Use AI to prototype or minimum viable product - use AI to create initial proof of concept drafts (of your ideas!) to test your ideas before you create them.
I occasionally do this before creating visuals and graphics for slide presentations, infographics, and diagrams. I’m not fully convinced I’m saving time on this yet, since my years of working in graphic design software services make it easy for me to mock up something closer to my vision in about the same amount of time it takes to prompt and refine an AI-prompted visual.
-
Use AI as a project manager - Use a tool like Claude Projects to house and help you organize your ideas related to a creative project.
I use this feature to organize the chapters and recall themes I want to include in the book I am writing, “A Dream Restored: Unlocking Latent Artistic Success At Any Age” [working title]
-
Use AI as a mentor, coach, or consultant - Use AI to learn, study, and quiz yourself on an area of your art that you want to hone. Or prompt an AI language model to ask you questions to pull out ideas, process roadblocks, and encourage you to get unstuck.
I taught a GPT all of the jazz theory approaches I wanted to learn and master for improvised soloing over ii-7, V7, I chords (a common harmonic building block featured in the majority of the canon of jazz standards). I then asked the GPT to help me create a practice routine, using the 80/20 principle to get the most development from focusing on the least amount of approaches with the most common usage. This prevented my precious practicing sessions from feeling scattered and broad and helped me master the most important methods that would get me the most results in the least amount of time.
-
Make art from AI-generated Art - The hallucinations, quirks, and happy accidents of creative AI tools can lead to interesting new art.
Not only can new styles and genres be blended, created, explored, and subverted, you can take it a level deeper and generate works based on AI-generated works.
For example, paint a collection of works based on really effed up AI-generated images or paintings.
-
Make art for an AI audience - Face it, Skynet is on its way, sucker! Why not sculpt a marble skull for your would-be T-800 pursuers to crush underfoot while you make a quick dash beneath the rubble?
In all seriousness, my favorite example comes from a fellow Berklee College of Music Alumni who uses AI to generate AI music for an AI audience. Cue an AI-generated remix of Rob Zombie’s “More Human Than Human.”
-
Make art that only you can make - Let’s face it, there are already humans that can do some aspect of your art discipline better than you. Their prose is more polished, they drum faster, they have better comedic timing, etc. That shouldn’t stop you from creative expression of your art and neither should AI art. You have to write the book only you can write, sing the song the way only you can sing it, show the world through your unique lens.
I’ll leave you with a quote I’ve generated from my human intelligence by training it on a sample quote from author Neil Gaiman’s famous “Make Good Art” speech:
“Husband runs off with an AI-assisted RealDoll? Make good art.
Leg crushed by the metal skeleton of T-800? Make good art.
Agent Smith on your trail? Make good art.
Tamagachi cat exploded? Make good art.
Somebody on the Internet thinks what you do is stupid or evil or it can be done faster by Dall-e 10? Make good art.
Probably things will work out somehow, and eventually time will take the sting away, but that doesn’t matter because we live in the Matrix. Do what only you do best. Make good art.