The Uncomfortable Truth About AI Creativity: What Creatives Refuse to See
What Creatives Get Wrong About Generative AI's Rise
Can AI be creative? what does that even mean? And if AI can be creative, how does that compare to human creativity, could we measure that even?
These are some of the questions increasingly arising from the ongoing development of generative AI that are harder to ignore as AI gets more capable.
There's much to unpack here. For example Arts vs Science: how do different disciplines look at the world? For the arts, the subjective and personal, what we feel, is often most important. For the sciences, it is usually more about what works, what can be measured, and what is more objective.
These disciplines and perspectives often remain segregated, and when they do combine, it's often a clunky compromise rather than a genuine synthesis, and the different worldviews of the arts vs science usually remain unchanged within those of each discipline.
Yet this is a fundamental and catastrophic error. I have trained (and have degrees) in both the sciences and the arts, and I feel this segregation of worldviews promoted by mainstream academia, has been very unhelpful, arbitrary, and counterproductive for improving our knowledge of the world generally.
Here we will explore how this segregation of the arts vs science worldviews to AI and creativity has led to considerable oversights and mistakes in how many creatives fundamentally understand AI and creativity.
AI Beats Humans on Creativity Tests
The MIT Technology review reported on a recent study that showed AI achieved higher than average scores than humans on a test commonly used on humans to assess creativity.
The article & study were careful to distinguish between passing the tests and if AI is being creative in a way that we understand:
The findings do not necessarily indicate that AIs are developing an ability to do something uniquely human. It could just be that AIs can pass creativity tests, not that they’re creative in the way we understand.
The test involves trying to come up with different solutions to a problem, specifically:
Researchers started by asking three AI chatbots—OpenAI’s ChatGPT and GPT-4 as well as Copy.Ai, which is built on GPT-3—to come up with as many uses for a rope, a box, a pencil, and a candle as possible within just 30 seconds.
While the results showed that on average AI performs better than humans on this task, the study also notes that the best humans do better than the average AI as well. The authors were also careful to specify the limits of their conclusions in this study:
While the purpose of the study was not to prove that AI systems are capable of replacing humans in creative roles, it raises philosophical questions about the characteristics that are unique to humans, says Simone Grassini, an associate professor of psychology at the University of Bergen, Norway, who co-led the research.“We’ve shown that in the past few years, technology has taken a very big leap forward when we talk about imitating human behaviour,” he says. “These models are continuously evolving.”
Another academic Ryan Burnell of the Alan Turing Institute also cautioned about the conclusions of this study:
The chatbots that were tested are “black boxes,” meaning that we don’t know exactly what data they were trained on, or how they generate their responses, he says. “What’s very plausibly happening here is that a model wasn’t coming up with new creative ideas—it was just drawing on things it’s seen in its training data, which could include this exact Alternate Uses Task,” he explains. “In that case, we’re not measuring creativity. We’re measuring the model’s past knowledge of this kind of task.”
Burnell’s comment while a common response, does raise questions about his own assumptions. What's the difference between an AI drawing on its 'training data' and a human drawing on its 'training data' to be creative?
Apparently, according to Burnell, it's creative if a human does that but not if an AI does it? Doesn't that feel a bit human-biased perhaps? a bit anthropocentric?
Where's the evidence to support Burnell's assumptions? That evidence might be hard to come by, given how little we understand how the human brain works and how creativity works in the brain, let alone how AI works. As a recent paper highlighted:
The neuroscientific study of creativity is stuck and lost. Having perseverated on a paradigm that is theoretically incoherent, there is little we know for sure about the brain mechanisms of creativity.
Burnell conveniently ignores the fact that human brains are also still essentially 'black boxes' we don't fully understand.
Leaving this aside, the study results do seem quite intriguing and suggest we should take the possibility of AI creativity much more seriously.
However, we should also be careful to distinguish as the study suggests that passing human creativity tests does not imply AI is doing the same thing as humans internally, or doing creative tasks in a way that is similar or understandable to humans.
Creative's Respond to AI Beating Humans at Creativity Tests
Of course, creatives are going to vary in how they think about and respond to anything, including how they think about AI and creativity, and how they would respond to a study like this.
However, I do find there are some common assumptions and frames of reference many creatives often use to understand AI.
With that caveat, I'd like to use as an example the response by Jasper Kense of the UX collective to this study in his article ‘Being creative in an age of genAI’.
Jasper makes some great points in his article, acknowledging the real impact that AI is having on the creative industry right now, and the understandable worry some creatives might feel about this and potential job displacement.
He's also rightfully cautious about what the implications may be as he says:
While the nascent stages of generative AI’s influence on the global industry are becoming apparent, the true extent of its impact is still unfolding.
He also makes a great point against naive assumptions of AI assuming it is some magic wand to solve every problem when it can’t actually do everything a human can, at least not yet:
Sometimes it is too easy to say: “Let’s use AI!” — Just like your manager might. We need to find the right tools at the right time.
When Creatives Confuse Feelings for Reality
However he then goes on to make several other points that seem more problematic, yet sadly common among most creatives.
For example, he makes the following assumptions about AI in the technical vs creative industries:
While it might be good to find patterns for certain industries, like data science, for creative industries it is but a tool. While it might be helpful for a designer to find 100s of variations of a toothbrush design, having one creative designer will give you the most unique design. A designer can put the generative capability to their advantage, leveraging visualization tools like Dalle2 and Midjourney. They can explore what the world has to offer. But to truly be creative, one would need to mostly explore their ideas and experience.
Firstly, it does seem somewhat hilarious he seems to believe that AI isn't a tool for some other industries, but it's 'just a tool' for the creatives. Newsflash Jasper: it's a tool for everyone.
Secondly, his belief that 'having one creative designer will give you the most unique design' raises many questions.
Who decides what's the most 'unique design'? It's one thing to use one subjectivity to create works of art, it's quite another to use one's subjective beliefs to make grand statements about all creativity.
The way ‘unique design’ is used here reminds me of
recent great article on Naive Realism and ‘self evident truths’ stated with total confidence, that on examination appear to lack much evidence or basis.This is perhaps one area where having some objectivity and dare I say, a little evidence can be helpful. The lack of it, perhaps is revealing.
Does Human Creativity & Uniqueness Even Exist?
The idea of human uniqueness and creativity may be more a myth than a reality..
We’re at the paywall. If you want to know what AI means for society with my weekly articles & podcast, with expert analysis of key AI developments and their impact, explained in plain english. Join today for monthly or annual subscription. Hope to help you get smarter about what AI means for society.
Keep reading with a 7-day free trial
Subscribe to The FuturAI to keep reading this post and get 7 days of free access to the full post archives.