Creative Uses for Generative AI

AI is everywhere, now, right? It’s available in several flavors.

Predictive AI, like when your email app prompts you to say “Dear Sam” when you begin writing, and “Let me know your thoughts” when you reach the end, is less controversial. Other built-in AIs, like those behind-the-scenes processes in our cars, refrigerators, HVAC systems, cameras, and phones, are taken for granted by many of us.

AIs may be brainlike in some ways, but they don’t look like this.
Image by Gerd Altmann from Pixabay

Generative AI

Generative AI, the kind of artificial intelligence that’s showing up in contentious ways, feeds on our human output to synthesize material that sometimes passes for human-made. Sometimes, it hallucinates, from being fed faulty or incomplete data. Sometimes, it’s eerily – and cheerily – helpful, as when a friend got expert lighting advice for a difficult photoshoot she was struggling with. And when its corporate owners use it to gather valuable consumer data from children, in the guise of a disembodied friend on their tablet, it’s just creepy, extractive, and deeply unethical.

Energy demands of generative AI can be huge, so there are major environmental concerns. It relies on training using billions of sources of existing content – and that processing comes at a heavy cost in both energy, for running the machines, and water, for cooling the processors. Predictive AI, by comparison, is relatively undemanding of energy and water resources.

When artists use AI

In a community zoom meeting with artist-organizer Andrew Simonet the other day, artists of many kinds shared a conversation around the ways we’re living with generative AI. A dancer used it to compile playlists of new music. A musician  verbally prompted it to create a song with a specific, unconventional, structure. Not only was it incapable of using the requested structure, they reported that the song was objectively just awful.

A writer discovered that a poetic metaphor of theirs had been scraped from JSTOR into an LLM where their creative turn of phrase will now live forever, uncredited for eternity. Someone suggested that using AI for logo design would be a permissible use, although an unemployed graphic designer might argue otherwise.

Another person recommended asking a threshold question before deciding whether to go with an AI or a human – Would I prefer to work with another human on this? If the money isn’t at hand, they suggested, consider using barter and other non-financial means of payment and continue to transact business with an entity who also has a pulse.

Racing toward the cliff?

Are we racing ourselves into oblivion? Possibly. We do seem hellbent on decimating the legal protections of US copyright law, which applies only to human-generated work. When LLMs (Large Language Models) train on noncopyrightable AI-generated content, look out. Even if your original human-generated work fed the LLMs, you’re out of luck in these scenarios. Print media guru Bo Sacks recently wrote clearly about this apparently inexorable trend.

Our own use of words is being modified by the grammar checkers and drafting assistance that AIs offer us. One professor at Simonet’s discussion told us they’d been encouraged to use AI in their graduate degree program, but now as teachers, they were told to prohibit its use among their students. The flat, affectless, uninventive language that AIs turn out will increasingly constitute the fodder being consumed for future generations. It reminds me of the project years ago illustrating what has become known as “generation loss,” for example, the degradation of an image as it was repeatedly, iteratively photocopied with an analog copier.

Where I use generative AI

Have I benefited from using it myself for creative work? Around the edges, yes. For instance, I remember when ChatGPT was newly available I prompted it to create images of the main characters in a novel I was writing. I did find that helpful, having visual references for the people in the story. When I ran into a gnarly problem with music tech, getting a plugin to work, I was less pleased. The AI adopted a friendly, helpful tone, which I appreciated; however, its advice was just wrong. Repeatedly. In the end, with a human on the phone at the company that sold me the plugin, we got things working properly.

I’m finding myself allergic to the ubiquitous invitations to indulge in AI-generated materials. Take web search. I understand that the top search result on Google, now its own AI-generated answer, is wiping out entire business models that relied on search for customers or clients. I’ve been skipping that answer on Google, scrolling down for the old-fashioned results, and, more frequently, using other search engines altogether. On my Vivaldi web browser (a more privacy-friendly, if Chromium-based, Chrome alternative), I use Startpage for search. (Other options include DuckDuckGo.) When I type in a search, I get a list of actual results, just like in the olden days.

When my email app pops up summaries of an email exchange, I ignore them. When it suggests an entire reply for me to endorse, rather than writing my own, I confess that sometimes I’ll read it quickly, to see if there’s anything useful. Once or twice, when I wanted a light, innocuous turn of phrase, I’ve adapted something it proposed. But generally, when I write to someone, I want to say what I intend, and that means actually figuring out the words that’ll accomplish that task.

Do I want it suggesting music or lyrics for my music writing? Nope. Do I want it offering me plot points or character profiles? No. Do I want it telling me where I left plot holes in my new manuscript? Maybe – until I remember that the price for that advice is providing my creative work to be scraped and added to a big LLM in the sky, or rather on distributed server farms around the globe. I’d prefer to stay out of that mess, as much as I can.

What’s next?

Artificial Generalized Intelligence is the next frontier, I’m told. Once that is possible, AIs will be a lot more sophisticated, able to create their own metaphors, for example. That means as well that the current fake-checking tools will be obsolete. It will be much harder to distinguish human-generated from AI-generated material in all media. In other words, we’re already at the slippery slope.

How do you roll with it?

I’m really curious what people are doing in their own creative – and other – practices, now that free and low-cost generative AI is on offer everywhere we turn. Do you use it? For what? Why? Do you draw a line beyond which you’re not comfortable accepting its assistance? What is that line? I’d love to know.

One last thing, for now, about generative AI: it’s inviting and easy. For now. As we already can observe, the backers of this technological revolution know how to bait the hook. Once we’re dependent, who knows how much we’ll be willing to keep paying for continued access. Our own AI subscriptions will become just another essential utility expense, like power, water, and internet. As the skeevy drug pusher in the playground always said to unsuspecting innocents, “First one’s free.”

Luna Project update: The third full moon, on 3 March, saw the release of Sleepy Moon. Awakening Moon comes next, on Wednesday 1 April. Visit my website or my YouTube channel for each new addition to Luna Project.

Thanks for reading!

~ Anne