On the day Richard Feynman died, he had a message scrawled across his blackboard: “What I cannot create, I do not understand.” As an expert in theoretical physics, he was referring to the idea that in order to truly comprehend a concept, or be effective with it, you must first be able to create that concept from scratch. While this may still hold in some areas of science and math, with everything that is happening in the AI space, I wonder if this remains true for creative work?
There seems to be a growing sentiment across the industry that because we can now achieve “good enough”, with less effort, we should settle for that. We should think small, short term, extract as much value as we can with minimum effort and hope for the best – without pausing to assess if that quality of output is good enough to distinguish us from anyone else with access to the same tools.
As creatives, we are now at a crossroads. While AI tools offer new possibilities, they also create a host of new challenges and requirements – as creative work gets easier to execute, it also becomes infinitely more complex to apply. Which begs the question: now that we can prompt our way to seemingly passable creative, does that mean we still understand it, or can be effective with it?
How is design getting easier?
Since the explosion of generative AI, the industry has shifted in terms of how we do creative work and in a sense, design has been simplified. One of these shifts recently occurred when Figma finally released their AI assisted design tool. Then quickly rolled it back.
If you’re not up to speed with what happened, here is a quick run through. Figma is the industry standard design tool for UX/UI designers. For a couple of years now, many smaller Figma competitors have been releasing plugins or end-to-end software that use AI to assist in the design process. Automating the creation of wireframes, components, assets, or even complete pages. Figma was obviously soon to follow. This finally happened at the Config 2024 conference, where they announced ‘Make Design’, an update that gave users the ability to prompt designs.
Shortly after, however, designers testing the tool found that if you prompted it to create a weather app it would design something that looked eerily similar to Apple’s iOS weather app. This gave the design community the ammo they needed to push back on the tool, amidst cries of copyright infringement, stolen training data and general anti-AI sentiment (understandably driven by fear of job loss). This resulted in the company apologizing and temporarily shutting it down. Don’t be fooled though, it will be back (already in beta with some users), or the company risks being disrupted.
In the meantime, we already have many other AI tools and features that make design much simpler. A couple of years ago, a UX designer would need a day or so just to put together a workshop board with discovery activities and the outline of some first-pass potential user flows. Now, this can be setup in minutes through Figjam’s AI tool. Even though the output is rarely usable on the first try, the fact that the basics can be set up quickly allows designers to begin their work faster and then adjust as needed.
Even from a user research perspective, at Create, we’ve shaved a huge amount of time from our insights process: going from an average of four days to transcribe user interviews and analyze the data, group the feedback, and propose key solutions, to a process that now takes less than an hour, using transcription and data analysis through AI.
We’re seeing this across the industry with tools like Wireframer for basic wireframes, Spline for generating 3D models (albeit, fairly useless ones for the most part),and Creatie.AI which allows users to add components using AI (which is basically just a smart search of a pre-built library). But still, it’s not hard to see that we will soon be prompting UI or entire platforms and the fidelity of the output will continue to improve exponentially. Not to mention the more obvious gains we’ve seen in concepting, asset generation, UX writing, and the huge gains in coding.
With many generative AI tools starting to amble out of the hype cycle ‘trough of disillusionment’ and begin their trek towards the much-coveted ‘plateau of productivity’, we are seeing rapid improvements in the speed and quality of output of digital work. But while these new improvements seem like they will make the lives of creatives easier, this double-edged sword also can also make design harder in many ways.
How is design getting harder?
Generative AI is built for system optimization, meaning these system will take whatever problem you put in front of them and solve it in the easiest way possible. Resulting in standardization of creative output. Which means that if creatives are being asked to implement these new tools, they also need to work hard to counter this large scale homogenization through new work flows and “forced” creativity using processes like Collaborative Chain of Thought Reasoning, or they risk the output of these systems creating a sea of sameness in which everyone’s work starts to look and sound the same.
We need only to look at what has happened to content on the internet, which is estimated to now be more than 90% AI generated. With increased pressure to deliver faster, copywriters are turning more and more to ChatGPT, without pausing to assess the work, set up proper prompting, or train custom GPTs to maintain brand tone of voice, resulting in posts across social media all sounding the same. I’m sure I don’t just speak for myself when I say that I’m starting to get tired of reading ChatGPT’s favourite words such as “delve”, “tapestry”, or “unprecedented”.
The perception that because we can now create faster, we can simply ‘do less’, is misguided not just from a sea of sameness perspective but also from a market competition perspective. For instance, in the past, with the advent of email adoption, office workers weren’t just sending the same amount of messages with less effort. Businesses suddenly scaled up communication. The organizations that didn’t adapt to this new way of working, sharing information, and connecting were left behind. I believe the same logic applies now.
This comes with a constant and steep learning curve that is evolving faster than any one person can keep up with. So, while the tools make it easier for creatives to do their jobs in some ways, the pressure and speed at which they need to deliver is increasing and the number of tools they need to have in their wheelhouse is growing. The increasing delivery speeds mean that it can be tough to find the time to master these new tools.
At Create, we manage this through a curriculum of trainings, workshops, lunch and learns, as well as individual growth-plan KPIs paired with time to experiment in hackathons. As a manager, I believe it’s the only way for us to succeed against increasing market pressure.
I’d even venture as far as to say that no matter what your job is, the growth of generative AI development means that learning new tools and skills will likely soon be a large part of your job too (if not already).
However, it’s not just how we work but what we make, too. The shifts in generative AI mean that our audiences and their requirements are changing. Today, if you are planning your meals for the week and ask Siri to order the required groceries, you will probably be left somewhat disappointed. Whereas recently we’ve seen OpenAI, Microsoft and Apple all announce updated AI assistants with a shift towards real-time interactions that are context-aware and in Apple’s case, can take action for you.
Now that AI will increasingly ‘ride along’ on devices and need to understand and interact with interfaces, there is a whole new set of user challenges that go along with this new kind of (AI) user. If not carefully considered, things can and will go wrong. For instance, Google’s recent AI blunder in which their model told users to use superglue to stick their pizza back together. What does that mean for UI design for instance? Especially as we know there exist new and undiscovered vulnerabilities in these tools.
New expectations from our human audience mean that we need to design completely new kinds of experiences.
Designers now need to spend time thinking about how generative AI could be embedded into platforms, which is a new set of things to design entirely. It’s a double-sided shift where on the one side AI will increasingly be embedded into platforms and on the other side users will also have AI assessing and interacting with these platforms alongside them or even for them.
Finally, designers also need to be thinking about the consumption and aggregation of these designs and content through search engines and frontier AI models that are serving this content to users in bite-sized pieces. Think of it as SEO but aimed at AI model aggregation. For instance, perplexity results, or Google’s recently rolled out AI based summarisations.
To add to this, these models are constantly shifting how they operate and consume or serve content. So, even if we can create and better understand something that was impactful yesterday, those wins can quickly be rendered moot.
How can we move forward?
There’s something truly disheartening about having just mastered a skill, tool or concept and then having AI make another leap forward and seemingly snatching it away from you. Change is scary. Especially when we can’t keep up with it.
You may have experienced it too. This kind of S curve of excitement of productivity where one second you are feeling empowered by an AI tool that can, for instance, break the dreaded curse of the blank page, allowing you to get started quicker. Then, moments later, you find out about another feature or tool that renders your input seemingly less important, making you feel lost or unwilling to try at all.
It’s not going anywhere either. I keep hearing claims from economists and even a well-known designer, that we are in a hype cycle of AI and that in a year or two, all of this will be forgotten as we go on about our lives. But that depends on how you define the cycle. Do we mean that there will be a stock market correction at some point and the value of AI companies might dip for a moment? Perhaps. Do we mean that many smaller ‘wrapper AI’ companies will go bust as they become consumed by frontier labs? Certainly, this is a winner-takes-all space.
But if you believe that AI will simply drift into the background and we will head into another AI winter, you likely have another thing coming. One thing we can be sure of is that change is inevitable and during these changes, the easier work will get, the harder it will likely be for many of us.
Narrowing down the scope to just the coming two years, how can we mitigate these challenges while capitalising on the opportunities? If you’re a creative, dedicate time to learning and if possible do it through side projects so that you learn faster, better and have the possibility of setting up side businesses that can capitalise on the intelligence explosion rather than waiting to be outworked by an AI agent. Instead of spending your time trying to fight the inevitable, why not learn to harness it? Find beauty in your new superpowers. Adapt and search for new ways to explore by tapping into that curiosity that made you want to explore new ideas in the first place.
As for executives, there are two fundamental ways to make money: by extracting value from a situation, or by adding value. So in this instance, where a large portion of the market is in a race to the average – trying to do more with less. Why not think about how your teams can generate new value altogether? Use this time of change to expand your impact and outperform those focusing solely on headcount.
Don’t cut your staff, instead double-down on them through education and process optimisation to stay ahead. It will be cheaper to build a culture of innovation and more efficient to up skill rather than re-hire with each new model release, as each new expert will come with a bigger price tag and you’ll lose your IP along the way.
Key takeaway
Despite these counterweights of gained efficiencies and shifting demands, one thing is undeniable – design feels exciting again. With the influx of new tech, digital work feels young, wild and unexpected. Like playing with Adobe Flash in the early 2000s, before everything was productised. Now, we can create new previously unimaginable brand experiences and have massive impact at scale.This marks a shift in our collective digital experience. Where ChatGPT was mainly focused on text-based interactions, we are now heading towards user experiences that are likely to feel more natural (if you can get over the awkwardness of asking your AI to do things for you in public). More translucent, more relevant, more enjoyable, and much more fun to create despite the challenges ahead.
But in order for this to happen we all need to get more comfortable with being adaptable. Leaning more heavily on our taste and intuition, and core skills such as curiosity, perseverance, agility, and of course, the ability to learn when to control AI and when to let go.
Thinking back to Richard Feynman’s much quoted message, it might be time for an update. Now that we can prompt our way to decent creative and new skills in the field are more about wielding these tools rather than understanding them in their entirety. We may instead need to settle for the idea that: “What I cannot control, I do not understand” and in the longer term we might even need to get used to the idea that we don’t understand very much about how anything works at all.
