Like many others, a few months ago, I started tinkering with various generative AI tools. These early experiments were mainly about exploring their creative and business potential. Almost immediately, one idea popped into my mind: AI-generated brand collaborations. So I started to build a little collab engine that could take two brands and a product and generate some marketing copy and imagery of what that imaginary collab is all about.
As soon as I started to see the results—and share them with others—I got very excited. It wasn’t just that this thing could spit out amazing-looking imagery, it was that it seemed to understand which brands were strong and which were weak. When you run a collab in the system with Hermes, for instance, the other brand must have a strong aesthetic, or Hermes will drown it out. Grateful Dead x Hermes? Yes. Financial Times x Hermes? Sadly, not so much. Somewhere buried in the machine was a fundamental understanding of what brands are and how they work.
As friends and I tried out more outlandish collabs, it was striking how on-target the system was. Sure, it sometimes produced strange-looking imagery or nonsensical copy. As a rule, however, the strong brands, regardless of scale, seemed to come out with better and more realistic results. As I thought and played more, I began to believe that this wasn’t just an accident. These Large Language Models (LLM) were trained on a massive corpus of data from the internet. They were effectively crunching the shared perception of humanity—at least the portion that contributes to the internet—into an understanding of how brands exist in our minds and lives.
Then it struck me: this is a lot like Brand Tags, a project I built 15 years ago. The tool—which showed people a random brand logo and asked them to type in the first thing that popped into their mind—was inspired by the simple idea that brands live in people’s heads. So when I compiled the responses into a tag cloud, with the most common words showing up in the largest font size, I had a snapshot of what brands really represented. It’s one thing for a Chief Marketing Officer or Creative Director to say the brand stands for refreshment; it’s a whole other thing for it to be the first thought of a consumer.
These memory structures sit at the heart of marketing. Great brands establish more connections and lodge them more deeply into the consumer and cultural psyche than their weaker competitors. In the end, at the heart of marketing and branding is the goal of establishing patterns in the network of neurons that make up the brains of potential buyers.
A decade-and-a-half after asking people for their tags, AI offers a whole new lens through which we can explore brands. After all, AI and machine learning are about using large amounts of data to interpret patterns and, in turn, deliver some form of understanding. Like our own brains, these systems utilize networks, frequently uncovering patterns that were invisible to human eyes. Unlike our own brains, however, these systems are statistical: underneath the output is hard data representing the computer’s understanding of the topic or idea.
All of this is to say that brands and AI/machine learning seem purpose-built for one another. The excitement—and fear—in the industry is palpable. People are beginning to recognize that this new technology will fundamentally shift the way we build and understand brands. And, like any change, that can be a little scary—particularly when the landscape is changing so quickly, and the industry’s understanding of the possibilities and threats is generally still blurry.
To me, it’s a lot more exciting than scary. I believe that the match between brands and AI is complimentary, and the best use for Large Language Models like GPT3, Midjourney, DALL·E, and the like, is to augment creativity, not replace it. I also think that buried deep inside these systems is a new level of understanding of what brands are and how they work. Just as Brand Tags reflected the perception of the masses, these tools have scoured much of humanity’s knowledge and creativity and quantified its patterns. And while they’re not without significant warts—serious questions of bias, for instance—they represent a new frontier for marketing.
BrXnd is positioned at that frontier. Sitting at the intersection of brands and AI, the goal is to establish a dialogue between the two sides that can help push both industries forward.
Beyond this initial collaboration experiment, I’ll be holding a conference in NYC in Spring 2023. The goal for this first event will be to establish a foundation: to discuss what’s possible today, how to make it happen, and where the future will lead us. We already have commitments from practitioners to share their projects and techniques, brands to share their experiments, and system builders to show where the technology is headed. I’ve also spun up a newsletter, appropriately titled BrXnd, that will explore the same themes.
If you have questions, ideas, or interest, please be in touch.