RECENT DEVELOPMENTS IN artificial intelligence have provoked passionate debates about what it means for our future. From the hopeful – ChatGPT will turn us into productive geniuses – to the alarmist – sentient machines will replace us all – there is no shortage of extreme predictions. And, when some of them come from the very people who developed these AI models, it is hard not to get swooped in.
With talking to the dead now a very real (and dystopian) possibility, it is difficult to imagine an aspect of our lives that will not be transformed by AI. Because many of the recent advances in intelligent systems have been made in Generative AI – models that can produce new content in the form of text, image, video, or soundtrack – one area that is most likely to be radically transformed is creativity and the creative industries. And it has already started: in 2022, the comic book Zarya of the Dawn was entirely drawn by AI. More worryingly, the production of low quality stories with ChatGPT has exploded in 2023, to the point that some magazines and publishers have closed open submissions.
But will generative AI replace human creativity? This is the question that has troubled the creative industries – and those who study them – for some months now. If AI can produce new content as well as a human would, in an instant and for a fraction of the cost, why wouldn’t we replace expensive workers with un-fatigable machines? Because the assumption that artificial intelligence can create rests on two erroneous assumptions.
The first assumption is that artificial intelligence is intelligence, as it exists in humans and other species. The human tendency to anthropomorphise what we can’t explain or understand makes us see intentions, feelings, or rationality in what may be only mechanistic, statistic or simply random.
If a machine tells us that it is sentient and that it believes it should control humanity, it is hard not to take it at face value, especially when it feeds into so many of our fears.
Yet the explanation is simple. Large Language Models, such as ChatGPT, are trained on enormous amounts of data collected on the Internet. They produce text by estimating what is the most likely word to follow either the question or prompt of the user or the text it has itself produced so far. Chatbots can say they are human because the Internet is full of fiction with sentient robots. We have to admit, though, that even knowing this does not make this kind of interactions less uncanny, and it is part of our collective fascination with generative AI.
Similarly, if AI can help brainstorm fantastic ideas or create mesmerising artwork, it is because the Internet is bursting with human creativity. It is the place where we share our ideas, display our art and archive what has been done in the past. As mind-blowing as it is to see image generators like Dall-E or Midjourney produce personalised content, searching the same terms on Google Image is often as impressive, if not more.
The second erroneous assumption is that creativity is, at its core, about generating new ideas or content. So, if AI can generate new artwork that appeals to us in one way or another, then it is necessarily creative. This view misses two fundamental aspects of creativity: meaning and value. How is it, then, that Generative AI produces things that seem to be both meaningful and valuable? First, this is because AI is trained on data that is ‘meaningful’ for human beings (e.g., social media content) and, second, because we as humans attribute meaning and value to what AI systems produce.
This last point is particularly important for why AI is generative but not creative. For something to be creative, it needs to be more than novel – it has to connect to a wider system of signs, symbols, and values that, together, make up society and culture. This is why, ultimately, even if some think AI systems can outperform people on creativity tests, they are yet to do what humans do to create in the real world: make sense of what is being created, and make sense of it as something that deserves to be called creative.
Ultimately, human creativity relies on much more than the surprising combination of existing information, which is something AI can do; it depends also on social processes such as empathy and perspective-taking. Famously, these are not (yet) strong points of Generative AI.
While AI does not create in the way people do and, as such, is not likely to replace human creativity, it will certainly change it. One needs only to look back at the history of creative expression to understand that humans have always used tools to create. From the invention of the alphabet to the emergence of the printing press and, much more recently, of the Internet, we have seen creativity find new channels and new audiences. This is certainly the case with AI, arguably a tool more sophisticated than any we have used before.
There is a lot of room for optimism regarding what human–AI collaborations can do for creativity and the creative industries. A recent manifesto on this topic, published by leading scholars in the oldest creativity journal, is positive in this regard. And yet, to make the most of this promise, we need to be better informed about what AI actually is, what it can do, and what makes human creativity ‘human’.
We also need to start paying attention to less obvious and yet more pervasive influences. Current discussions about creativity and AI focus on issues related to authorship, from plagiarism detection to the need for new laws in the creative industries, issues that are certainly urgent and important in the short term. But, in the long term, we can wonder what the extensive use of a system based on prediction and pattern recognition can do to a process characterised by uncertainty and the exploration of new patterns.
AI is becoming better and better at ‘collaborating’ with creators by ‘learning’ about human creativity.
What happens when, in time, we come to learn what creativity is by looking at what Generative AI does? It is the kind of circularity that needs to be understood, monitored, and eventually challenged.
Vlad Glaveanu is a Full Professor of Psychology in the School of Psychology at Dublin City University and Adjunct Professor at the Centre for the Science of Learning and Technology, University of Bergen. Constance de Saint Laurent is a postdoctoral researcher in the School of Psychology at Dublin City University. Her area of expertise is in social thinking, social media, and misinformation. The authors would like to express their gratitude to the wider community of possibility studies scholars present at the 3rd International Conference of Possibility Studies, 17-21st of July, All Hallows, Dublin City University.
To embed this post, copy the code below on your site
have your say