Let’s address the AI elephant in the room. You’d never hire someone this mid

We all marveled at the power of ChatGPT 3.5 when it came out in late 2022. Not long afterward, ChatGPT 4 showed up (though behind a paywall) and proved to be far more powerful. It makes any reasonable person wonder what the next versions will look like.

If new versions of ChatGPT roll out like other technologies we’re familiar with (like iPhones), then won’t the machines be running everything soon? It would appear that Open AI’s vision of AI “doing all economically viable work” will come true before we can utter the words “copyright infringement.”

But, as it turns out, gen AI is not on an infinite upslope.

All the best evidence points to it being on an S-curve. So, breathe some cautious sighs of relief marketers, novelists, creators of all stripes. Gen AI is gliding on to a plateau.

Yes, there is an AI hype cycle, and no, it’s not ALL hype.

Don’t fall for the hype

You could be forgiven for believing that gen AI is on an infinite upslope because people keep talking about it taking over the world. And come on, admit it – you fell for the hype cycle at first. I did. I’ve been working with AI/ML for decades and these new models were something different, something we hadn’t really seen before.

We made art. We generated blogs. We carried on conversations that obliterated the Turing test. We cranked out B- creative ideas and poems (well, rhyming verses, at any rate). And while it is true that we never need to face down a blank page ever again, the promise that these technologies would *change everything* was a bit too much.

Once I had caught my breath and thought about it as a software developer, things looked different. Yes, these tools are really interesting but not in the way that you think.

But if you fall for the hype, you risk falling flat.

The sea of sameness

At a hand-wavy level, both Transformer models (like ChatGPT) and Diffusion models (like Midjourney) are statistical models that predict the next most likely token. For GPTs, that means words. For Diffusers, that means pixels.

If you’ve worked with these models in their various shapes for a while, you’ll start to notice a sea of sameness. The text generated feels flat. The images are striking at first, but eventually, the novelty wears off as the images have the new New Aesthetic. The machines are making average things: because that’s what they are designed to do. They aren’t witty, they aren’t funny, and they aren’t insightful. They are algorithms that spit out the next-most-likely-thing based on the data and settings you use.

And the “intelligence” in artificial intelligence is a bit of a misnomer because you would never hire someone this mid: “Draw a room with no elephants in it.”

And yet, the ability to get four nice drawings of four nice rooms with elephants in them – in a matter of seconds – isn’t worthless, far from it. For a creative mind, possibilities abound. You can speed up the process of ideation. Package up a new tool or integrate a shortcut into an everyday process. Expand into new outputs and do your damnedest to amp up the quality, such as videos made with Sora. Or even (jargon alert) put utterly unique, new-to-world personalized content in places and at scales that no human or team of humans possibly could.

Just remember that having AI or using AI doesn’t make you special. Doing something creative, or soulful, or humane, or useful with it makes you special.

The good news for marketers about AI’s plateau

And that’s good news because here’s the truth of this moment (looking at you, my fellow marketers): it’s the right time to stop and figure out where to go and what to do with the AI we have right now.

The big takeaway for gen AI is an old take with new urgency: never bail out on the full creative journey.

When we stop to take a breath as we approach the AI plateau, we can gather our thoughts to ask some very real questions:

• If the new baseline is what a Diffuser or an LLM can spit out with zero effort, where’s the new creative bar? (Hint: much higher)

• If we have a higher level of mediocre output that we can reach in a matter of seconds, are you going to settle for it? Before you answer, consider this other question: have you ever settled for mediocre before and gotten away with it for very long? (Hint: you’d better not have, and you’d better not now)

• Are you going to replace the majority of your teams with AI or hire marketing agencies who are doing that? (Go for it–but only if you want the same kind of work anyone with commonly available AI tools can do)

AI is best used, generally, when it removes pain points and creates new opportunities for things that weren’t possible before. Gen AI is best used, specifically, for the same reasons.

No one is looking to hire a Photoshop agency. Why would you want to hire a Midjourney one? Tools are tools. Optimizations and opportunities have and will continue to come. Integrate the optimizations, capitalize on the opportunities, and trust creative people to come up with new types of work we’ve never seen before.