Meeting at the Middle Between AI and Creativity

By Michael Dobell, CIO, Jam3

When the potential of a new technology starts to become evident, we as an industry seem to make the cognitive “skip to the end.” The natter on LinkedIn these past eight weeks began with doom and gloom for the creative profession as more and more sophisticated uses of Midjourney, Stable Diffusion, Dall-E, LaMDA and ChatGPT3 emerged.

It’s not at the purported “end game” where opportunity lies, but rather the middle. Most of us think in a binary way, fast-tracking to an end state that’s easy to envision while ignoring the messy middle that lies between. But in doing so, we neglect a good opportunity to assess and prepare for risks—much like the doomsday prepper who fails to consider how to build a better local community today.

A creative losing their job to an AI may be easy to envision and worry about, it’s not the likely outcome of these tools. Rather, the more complex questions about how we build, value and claim ownership of creative may prove more pernicious, and it’d behoove us to start asking them now as we pick up signals into our AI-assisted future. It’s the steps we take today, right out the gate, that will determine what that end state looks like—so we need to get them right. Let’s slow down, take stock, and think on how to build a brilliant future for the creative industry.

Glimpses of opportunity

In the twenty aughts and twenty teens, we developed some pretty magical new powers through a thing called “Search.” Over the decades we acquired a mechanical sympathy for the machine, able to pull precise bits of knowledge and integrate it into our creations, all without much knowledge of how it all worked. It was like when Bilbo found the ring: we quickly understood its power, but only over time came to realize the risks.

There’s a similar pattern today with large language models and AI image synthesis, though oh so much faster. Our studio ran a hackathon on the creative applications of Midjourney in November; just six weeks later, a creative lead developed a proposal using entirely AI-generated art and copy made more compelling with assistance from ChatGPT3. Our tech architects built a GPT3-powered Twitter bot for the FX series Atlanta just last month. And how about selfies from a time traveler synthesized from a whole suite of generative AI tools? That’s some great comedy right there.

Even people outside the creative space are jumping on these tools, not just for fun (“Rembrandt-style portrait of a shark riding a motorcycle, please…”) but to craft great sales emails (“Introductory email to a franchise restaurant owner sharing the features and benefits of our software booking solution, please…”). With the general population now using these tools and supporting services like Prompt Hero, mechanical sympathy for these tools is quickly emerging.

We’re also learning in glimpses of bizarre new economies. DoNotPay’s GPT3-powered bot will negotiate a better rate on your Comcast bill (and maybe exaggerate the severity of connectivity issues), suggesting a future where bot armies duke it out. But is there value creation here, or just sand in the gears? To move towards an answer, I’d like to look into a very near, possible future of what the creative process will look like when augmented by AI.

Glimpses at risk

Sammy, a freelancer working on a project for a startup client, boots up a blank Figma file. They begin making initial logo sketches, and after days of ideas bumping in their brain, it’s almost a relief to get them out. Logo Hero, a generative AI, offers variations that are helpful, sometimes brilliant, often crap, and easily excluded. From a small selection, Sammy requests new variants. The dumb brilliance of the machine—they riff together a while longer. Sammy sometimes wonders, “Who’s leading who?” But it doesn’t really matter.

Narrowing the field to a few versions for tomorrow’s client call, Sammy selects their favorites and inputs a verbal directional prompt into Tagliner, a tagline-generating tool. But only the lines with the lowest license fees will be okay with this client. Sammy knots their brow at the realization: all the best taglines and logos were already generated and trademarked en masse by Madison Avenue’s bot armies—hundreds of millions of assets and copy locked away behind one heck of a licensing fee. It’s the classic patent troll playbook, now amplified by AI snapping up IP in seconds. And it’s been like this ever since the removal of barriers to creative at scale.

To glimpse at the future, reflect on the now

In this possible future, the danger isn’t that AI took away anyone’s job. Rather, our 20th-century legal models and regulations haven’t adapted nearly as fast as the ability to generate and copyright infinite creative material. So, while it’s tempting to skip to the potential end state of a new technology and the new problems they may bring to the table, it’s worth taking a step back to consider how they’re more likely to amplify today’s existing ones. The good thing? There’s still time to solve them—so let’s get to work now.