I have mixed emotions about the current hype around Generative AI (GenAI). As a freelancer in the AI space, I’m excited to see the breakthroughs of the last decade going mainstream. Thanks to ChatGPT, imaginations are catching fire. It’s great to see so much public interest in new roles like Prompt Engineer, AI Auditor, and AI Ethicist.
As an academic researcher using AI to do semiotics, however, I know how difficult it is to get the theory, data, engineering, and ethics of Applied AI right. I don’t see that being part of the GenAI brand book though. I dread the fear, uncertainty, and doubt – the ludicrous promises – that lie ahead. And I’m not just talking about the small-time opportunists. OpenAI’s curt and careful blog post updates on ChatGPT signal to me that the main ethical and IP blitzkrieg is about to start. Reportedly, OpenAI lost around USD 540 million last year while training ChatGPT. Scheming about monetization of the “research preview” must be truly Medicean.
Branding seems twice cursed
The AI upheaval wouldn’t be a problem for branding, of course, if the hype was just hype. Like other creative industries, branding is on the doorstep of a forced evolution. Branding seems twice cursed: GenAI will repave consumer consciousness as well as the means to reach it.
Amidst the coming deluge of AI sensations and sales talk, many branding pros and brands will fail. Most will fail despite having a bold AI strategy. Why? Partly because everyone will soon have a bold AI strategy, but also because bold to many means big.
Brands will fail simply because they believe plug and play AI can deliver more than it really can.
There’s little doubt everyone in branding needs to invest in AI. But now that the impact of GenAI is obvious, many branding pros are stuck in the paradox of clarity. They’re finally ready to get serious about AI because they can see the potential advantage (agile text and image generation, research automation, new horizons in consumer access). But so can everyone else.
Be what AI is not
There are no norms. Nobody understands the future. Politicians and regulators are years behind. I read the AI papers. I talk to AI academics and engineers. Nobody knows. Very few in AI even know what semiotics is. Even in the somewhat adjacent field of natural language processing. So, to paraphrase Richard Feynman on quantum mechanics, if you think you understand AI and branding, you don’t understand AI and branding. Take this as rule one.
Embrace the very skepticism that is currently trained out of AI.
Here’s the paradox. Rule one isn’t a reason to languish in despair. Properly viewed, it’s pure empowerment. For if AI is about to change things in big unknown ways, there’s no choice but to act. To transform. And unique agile strategies can turn danger into opportunity. There’s not going to be a suite of products that level the playing field. So rule two is to put rule one in motion – trying new things with AI while keeping risk sub-apocalyptic. For creatives who can surf uncertainty – who can enjoy the distractions of AI while studying the ways others are distracted by it – skepticism in motion should be a familiar skill.
Three ways to put skepticism into practice
Firstly, skepticism exhorts you to invest in proprietary AI systems only for very specific purposes. But what mix of money and time is optimal? Let’s take an example at the very pointy end of AI that branding pros might use. Should you spend 20 or 30 dollars a month on a service like ChatGPT Plus or Anyword, or use their free versions? One reason many AI firms offer free subscription levels is to gather user input to further train their models (we’ve already seen the free ChatGPT 3.5 give rise to the subscription ChatGPT 4.0). I recommend not playing this game. Save time, not money. So, yes, pay for subscriptions at the leading edge. You want your creative time enhanced with the best tools, not hamstrung by obsolete ones.
Just don’t get comfortable. If ChatGPT or Anyword don’t fit your next project or larger vision, should you invest 20 or 30 hours a month trying to make them? No. Invest that time in exploring better fitting options. Drive AI, don’t let it drive you. Big tech products are designed to be sticky: the more time you invest, the more you’ll be stuck – not just with the subscription, but with a cap on results that comes from being part of the crowd. It may seem like you have few options, but a tsunami of products and collaborators is arriving in the AI space. By putting time into finding what fits your unique creative offering and data (rather than making yours fit theirs), you stand a better chance of building a solution grown from your own skill set – a unique selling point. It may not be everyone’s cup of tea, but the winners in the AI-driven branding future will likely have exceptional stamina for pulling diverse partners and technologies together for short collaborations, rinsing and repeating until a breakthrough project is achieved. Their data and trade secrets will be their own.
Good policy, therefore, is to play at the margins with a view to developing your own bespoke mix of AI with other players at your scale.
Put buying killer one-size-fits-all systems off until they become an obvious, state-of-the-art proven resource essential to everyone.
The current hot property AI agents built on GPT-4 and LangChain like AutoGPT may go that way – but they’re not there yet. Branding is too complex for someone to invent a machine with one big button that solves all problems. A change in the AI landscape will occasionally force you to adopt a de facto standard piece of tech, but when it does just look for novel ways to apply it. The creative zone with AI is in taking the latest tech and adapting it to a new task (in fact, in technical application within AI it’s called “transfer learning”). If you’re invested in the idea that someone else has figured everything out and can sell it to you in a black box, you will have little ability to pivot and invent something fresh and unique.
Secondly, skepticism in motion improves your relationship with jargon. You don’t need to know the mathematics behind temporal word embeddings or generative adversarial networks. But you need to have the functional possibilities they afford clear in your mind. Without that high level clarity you cannot engage your creative abilities to combine them into something new. Ask consultants and vendors for intelligible explanations, of course. But be more active than that. When building a narrative that includes AI components, insist on academic, scientific evidence that reconciles technology with outcomes. Otherwise, you will invest in faith more than is wise in a secular marketplace.
Welcome honest ignorance. If those you consult or look to partner with tell you something is uncertain or impossible to know without further research, be ready to treat this as a positive sign.
For example, I’m often asked if shifts in the mathematical representations of words in AI models translate to cultural and behavioral change. I usually answer, “That’s what we need to develop a thesis about.” Much current branding research is predicated on the assumption that linguistic models can capture changes in thinking. But anyone who gives a blanket “yes” to any ethnographic question is too much a salesperson or engineer.
Thirdly, skepticism in motion dictates you grow your AI tools and methodology from the ground up. Heeding the comments above, harness the hype: meld AI into your pitches and projects. But in practice, tie it to very specific scientific aspects and be ruthless about where you think it doesn’t fit. If even your most tightly focused use of AI falters, be scientific about analyzing the results and revising the approach. Privately. Be ferocious about protecting your trade secrets and intellectual property. If a transformer-based topic model works wonders modeling advertising touchpoints and sales of cola in summer in Sweden – that’s gold, your gold. Even more importantly – be coy about what doesn’t work. Uncertainty will be costing your competitors money and results. Don’t help.
AI experiments might seem purely technical but they are business intelligence.
The goal should be to build a proprietary AI system that reflects your hard-won branding insights and process expertise – not your desperate adaptation of logic built in someone else’s software.
Holding two opposing ideas
Enjoy clarity where you can but learn to think in probabilities. Take calculated risks. And if they fail, take lessons. Pivot.
A decision in the face of certainty isn’t a decision. Those who wait for certainty are destined for mediocre results at best.
Optimal investment in AI boils down to holding two opposing ideas at once – what F. Scott Fitzgerald saw as the test of first-rate intelligence. You can’t understand everything, so you can’t fully believe anything. But you must act as if you do. Embrace the blending of AI and branding. But be wary of big steps or grand narratives. Be eclectic. Small steps; point solutions. Scaffold on your successes. Skepticism. Action. It’s a human approach. An agile approach. A steerable, presentable, ethical approach. And one that allows constant creativity through reinvention of tools.
Cover image source: Андрей Журавлев