There’s been a lot of concern and hand-wringing about some of the new machine-learning tools like Chat GPT, Dall-E, etc potentially being able to replace artists, designers, copywriters, musicians, etc.
So I decided to see what the fuss is about.
What would happen if I tried to re-create some of the most iconic ad campaigns of the last fifty years?
I chose “Got Milk”, “Just Do It”, “Think Small”, and “Where’s the Beef?”
I didn’t try to make shitty or goofy ads on purpose– I gave parameters to the machine learning tools that described the end goal the way a good Creative Director or even well-meaning client would provide.
For instance, for the “Think Small” copy I gave the following parameters: “Give me a one paragraph essay under 200 words describing our car as a slow, ugly, not cutting edge, but very stable and reliable performing vehicle. Use very short sentences and include “our car” in every sentence.”
For the “Just Do It” ad I asked “Generate a viral simple catch phrase that motivates you to do anything with no limits”
Image requirements were on the same lines– a specific shot, a certain subject within a desired age range, and an “emotion” that fits with the tone of the ad.
So some of my key takeaways:
- good design means picking very specific shot angles, closeups, etc which is hard to do with these tools
- It’s really, really difficult to get an image that “feels” right, even if it’s what’s asked for (“athlete in sportswear at the edge of exhaustion pushing through to victory”). Since visual design relies on the PERFECT image with the PERFECT copy, the results are…. challenging.
- copywriting is possible to get kinda close, but EXTREMELY difficult to get right, or even passable. I could get the right “vibe” by re-generating with new parameters but by then I might as well have written it myself
- the “Think small” ad in particular is stunningly well designed. It was very difficult to resist the urge to adjust the layout to emulate Helmut Krone and Julian Koenig’s masterful design and copy. The way they use empty space in both the images AND the text could be a design class in itself.
So what’s the verdict?
I’m not super worried about losing my job (if I had one, I mean…)
Getting machines to get the “vibe” of an idea is really, really tough since even humans have difficulty with it, even if they’re not Missing Missy.
It may be harder for new talent to build a working portfolio since low-quality, low paying clients will be less likely to hire a designer when they can “just type something into the computer and get a design”… but those clients aren’t gonna hire talent anyway.
Good clients can tell the difference between barely good enough and fucking amazing, and the smart ones are always willing to pay what it’s worth..