Discover more from Better by Design
Why I changed my mind about AI for creative work
First impressions from the week AI took over my feed.
This was the week that AI took over my feed.
The week where its power, potential and impact on creative work started to sink in.
Guess it blew in with the fall air.
First, it was Pieter Levels smart execution of an interior design staging app built on Stable Diffusion (mentioned in last week’s post). Then, it was Derek Thompson’s podcast episode “Why The Future of AI Should Terrify and Thrill You” and Evan Armstrong’s essay “Who Wins the AI Value Chain?”.
I’ve barely scratched the surface of what will surely be a long, deep dive into this topic, but I wanted to share what’s piqued my interest so far. So today I’ll highlight points from Derek and Evan’s work that got me thinking and then cap it off with a quick thread co-written by me and the AI language model GPT-3 .
“Oh great, it’s actually good”
Now, if you’re a bit of skeptic like me, your initial instinct might be that we’re just entering another tech hype cycle. It’s an easy out, but falls apart quickly once you look under the hood of what’s happening.
Honestly, I think I held off investigating AI more deeply because I didn’t really want to reconsider everything I knew I’d have to if it turned out to actually be good. But pretty much as soon as I dug in, I had a bunch of ideas for how I might design experiences based on the current foundational tech… so, damn… I guess it’s actually good.
As Evan puts it in his article:
I entered this period of [AI] obsession predicting that I wouldn’t find much of anything. The last time I did this with crypto’s hype cycle, I [found]… only one use case I loved. I expected similar failings with AI hype. Instead, I found the opposite. There is real, immediate merit to the industry.
As a product designer, the new possibilities are riveting. But as a creative professional, it’s nerve-racking to think how those possibilities will fundamentally change how I engage with and perform my own profession.
I sometimes joke that my goal when creating design systems is to “design myself out of a job”; a tongue-in-cheek take on using systems to free myself up to focus on the bigger picture. I always meant it as a joke, but this new tech, well, it might allow me to literally deliver upon it.
“But I thought design was safe…”
To this point, when I’ve talked to other designers about AI the resounding response is something along the lines of: “computers aren’t good at creativity, they’re good at manual tasks, so we’re safe.” To some extent that’s true, but I think we’ve underestimated how much of our creative output is actually powered by systematic tasks carried out in rapid succession.
While I personally don’t think AI is coming for emotionally resonant, creative art anytime soon, most day-to-day, professional design work doesn’t fall into that category. It is creative work, but it’s not art. Instead it’s more like creatively executing tasks in a media format based on a developed and trained sense of pattern matching. And it turns out that executing routine tasks based on identified patterns is what AI is incredibly good at.
Legitimate, routine work for a designer turns out to be well oriented for “creative” execution by a pattern matching beast like a supercomputer.
I need to create new imagery within the bounds of my brand identity to use on my website and blog posts.
I need to apply existing patterns from my product design system to construct a CRUD interface for a new object in our application.
Both are important tasks, but ones that I think we could certainly train these new AI models to do. And honestly, I think we’ll end up happily delegating those routine jobs to a computer sooner rather than later.
Why was I so blind to this possibility?
I think it’s likely a combination of a couple cognitive biases.
On one hand we have the “Endowment Effect”, which leads us to tend to overvalue things that we own. And on the other hand we have its close cousin the “Ikea Effect”, where we overvalue things we have helped create.
As creative professionals our creative output can feel like our baby; we brought it into this world and we are inseparable from its existence. It’s no wonder we feel like our work could never be replicated by a machine; we’re way too closely attached to it to be objective.
But in reality, we’re learning that a computer can break down, capture and quickly repeat the basic steps of the creative process much better than we ever anticipated. It might not overtake the creative birth of truly original ideas, but it could easily step in for the iterative process of designing on top of an established foundation. And that work actually represents a lot of what many designers currently do in their day-to-day.
Reverse engineering our own evolution
A counterintuitive point from Derek’s podcast caught my attention.
He and his guest Kevin Roose talked about how our progress in AI reflects what is basically the reverse engineering of our own evolution, from least evolved traits to those most deeply-seated.
What caught me off guard was that my initial impression of “least evolved” was entirely backward. My instinct was to think of creativity as the most sophisticated of all human faculties. The latest and greatest of human evolution. But, biologically speaking, being the latest and greatest actually means that those traits are the youngest and least evolved of all our abilities. Relative to something like keeping your balance while moving down stairs, those creative faculties have had far less time to develop.
So, when you look at what we’re able to capture with AI, it makes some sense that the first pieces of our own evolutionary makeup that we’d be able to recreate would be the ones that developed most recently in our history, not those that originated at the dawn of our species.
First impressions on my GPT-3 use
So far I’ve been using GPT-3 in a narrow capacity; basically like an evolutionary Google. I ask it questions or prompt it to return research quickly in a format that’s more immediately translatable to the writing I plan to do. The model is very good at outlining “key points” about topics, which is a useful discovery aid for me even if its suggestions aren’t really in line with what I was thinking or meaning.
Unlike Google, GPT-3 won’t suggest multiple sources for you to check out in order to make your own conclusions. Instead, it jumps straight to making a conclusion for you regardless of whether it’s right, wrong or accurate to the way you were thinking about the question. In that way, it functions a bit more like asking a knowledgable friend for their point of view on a topic; they’re likely to make some good points, but it’s still just their unchecked POV. So keep that in mind and proceed accordingly with any results you get.
I’ll conclude by sharing the response to one of the many prompts I threw at GPT-3 in the past week. Of all the answers I got, I thought these five points in response to “What should designers understand to be successful?” hit pretty close to home.
Welcome to the era of a computer giving you good design advice. Prepare to be thrilled and terrified.
Signing off 🖖,
5 things designers should know, according to AI.
(Headings by GPT-3 🤖, descriptions by me 🤓)
Know your audience
Every design has an audience.
It doesn’t matter if it’s an audience of one, or one billion.
Your success hinges on creating something that resonates with that defined group.
Know your product
Whether you're designing marketing materials to sell a product or designing the product itself, the better you understand the domain, the better you can craft it.
Know your constraints
Constraints aren't constraining, they free you to focus.
To try to design without them is to set yourself up for frustration and, likely, failure.
A lot of design opportunities unlock when you simply identify a good set of useful constraints.
Know your design process
Know your design process. Not just the design process.
Learn from the best and then riff on what they do to find what works best for you and the current environment you’re designing in.
Know your tools
Ahh tools. The heavy hitters of the social media scene.
Extremely important insofar as they allow you to express your ideas quickly and accurately.
Not so important in terms of which tools you prefer to get the job done.