Good Trouble for AI-dman
You can have it fast. You can have it cheap. You can have it be good. But you can’t have all 3 what happens when you can have all 3?
What’s new in AI and marketing?
This happened this past week:
GetDirty, a wetwipe brand, was able to create an influencer social ad out of AI based on a real model. The cost? $100 and it took minutes to create.
The model, Ariel, is a freelancer who submitted herself to be modeled by companies like Arcads who in turn helps others create ads out of AI actors.
GetDirty, for its part, admits this was done for publicity and has since hired Ariel directly (you can see the side-by-side).
Is it perfect? No but it passes for real enough, especially for social.
But this is only good for social now, right?
Not for long.
OpenAI’s text-to-video Sora debuted this February to a handful amount of visual artists, designers, and filmmakers.
Native Foreign’s CCO Nik Klererov (a LA-based agency and production company) is one of them and has been releasing videos of his experience to date.
Nik’s looking at this as a creative, even saying I feel like a kid again. But how does he see Sora for the near future?
At the very least, Sora is an iterative tool to workshop the lighting, the vibe etc to what this could be.
To use as B-roll or establishing mood shots, that could end up just being Sora shots with some clean up.
OpenAI plans to open up access to Sora later this year.
So AI can copy what we feed it but it can’t ‘sell’ like a human right?
Sure.
AI Is Becoming More Persuasive Than Humans
A new study reports that a personalizing LLM* was significantly more persuasive than humans within an online setting, by more than 80% (p<0.01).
That is, when faced with an LLM that has access to demographic information allowing it to personalize its argument, humans are 81.7% more likely to agree with the arguments when compared with a human adversary.
*Large language model or generative AI
The study was centered on ethical issues such as race, and not choosing between laundry detergents, but it’s not hard to see its application for marketing.
Other research suggests humans already prefer AI-generated content if we don’t know it comes from AI (being told the human copy came from a human only gives it a small boost).
And AI can ‘read our emotions’ (see Hume.ai) which means it can alter its message based on how we are reacting in real-time. How far off are we from bespoke ads tailored to your mood?
The Search for the Holy Grail of Personalization
73% of marketers say personalization is important — only 35% of marketers feel that customers currently get a very personalized experience with their brand. — Hubspot 2024 State of Marketing
If you have to guess where it’s headed, it’s in the word Personalization. The promise of each experience is tailored to the individual and therefore is the most useful, engaging, and persuasive.
What’s holding this up right now is
1. tech isn’t quite there
2. the data isn’t quite there
3. infrastructure (microchips, power, regulation)
But as we can see, the tech will get there. Remember we’re only one year from Will Smith eating spaghetti and now we’re here.
Right now, we can mimic anyone (even Ryan Gosling) to a pretty convincing degree, if you don’t look too hard. Remember, most of us are watching on small devices while scrolling. Now when we get to the point where, in real time, Ryan can become Margot Robbie for someone else remains to be seen.
As for signals, we currently rely on a combination of first-party / third-party (ie cookies) along with modeling to ‘know’ who you are. Third-party is likely going to go away, but that doesn’t matter because the internet has become incredibly centralized. Social isn’t just social, it’s news, it’s search, it’s one-stop shopping literally. It knows everything about you because you feed it whether you post yourself or not. It’s primed for engagement. And now everything will have chatbots, LLMs to which you feed even more information into it. They’ll know enough.
Regulation and processing power are the two biggest speed bumps right now. Even if everything created by AI for you was mandated to have a disclaimer, I don’t believe people will care. Why? Because it doesn’t bother us to see the paid endorsement next to influencer posts. It doesn’t bother us that Instagram serves us up an ad about the thing we were just talking about. It doesn’t bother us that everything we see is fed to us by an algorithm.
AI will be just one more thing tailored to sell us.
And as for chip power, go watch Amy Webb’s SXSW talk. They’re figuring out how to turn living cells into microprocessors, acting like ‘little brains’. Insane — we are turning into the Matrix.
As for what this all means for the ad industry, there are POVs that run the gamut from excited (see Nik above) to we’re all doomed.
In my experience, the more targeted the ad gets the more the idea becomes about you. Emotion through ‘being seen’ and talked to directly. The limitation typically comes from having the right asset (product, talent, message) to match.
As we sit here today, it takes humans to do all of this. What happens when it doesn’t? How soon til it’s good enough? And for how much longer will there be enough on the broader Idea-driven side to ‘keep the lights on’ for ad agencies?
You can have it fast. You can have it cheap. You can have it be good. What happens when you can have all 3?
