Are computers capable of creativity? How do we evaluate whether they are? And how can we ensure their outputs, creative or not, can be trusted? Dr Anna Jordanous is working within the Artificial Intelligence and Data Analytics (AIDA) research group in the School of Computing to help us navigate the widespread adoption of AI in society and bring us closer to answering these fundamental questions.
What can we gain from examining the link between computers and creativity?
Computers have become an essential part of our lives. We constantly use computers to help us find out information online, guide us from A to B when we’re lost, document things we are doing at work and, increasingly, generate impressive creative content. This begs the question – is this just ‘mindless’ processing or are computers capable of creativity?
In computational creativity, we explore to what extent computers can be creative. The goal of computational creativity is to model, simulate or replicate creativity computationally. This means we try to get computers to do creative things and, in turn, learn more about our own creativity. I’m especially interested in how we can design models which help us understand human creativity better.
How is your own research helping us understand creativity?
In 2016, I led work to identify 14 different components of creativity: aspects of creativity which are key to understanding how creativity can be defined. For this, we analysed the words used to define what creativity is, taking into account lots of different perspectives. As well as helping us understand our own creative practices, the model can be used by computer scientists to identify what aspects of creativity should be programmed into a computer to achieve a particular task.
For instance, as an amateur musician, I applied the model to a program I had designed to improvise music. The program didn’t perform well compared to similar programs but by speaking to human musicians, I was able to identify which components of creativity were most relevant to the improvisation process. That helped me work out what I would need to focus on programming into the system to improve it. Of course, that poses the question of how you program those elements –such as social interaction, intention, emotion and musical competence – into a computer!
Since the explosion of AI technology in early 2023, using computers to generate creative outputs has become almost ubiquitous. How has this transformation impacted on your field?
The widespread accessibility of AI tools has indeed been unprecedented. What’s interesting is that AI, as a discipline, has been around for decades, but recent developments, including the availability of accessible transformer-based AI tools such as ChatGPT, Bard and Llama, mean that anyone can now leverage powerful AI models for creative tasks.
Actually, this accessibility only makes my research more important, as there’s a more pressing need for us to understand and critically examine the implications of this technology becoming universal. While I may not have access to the vast amounts of data that companies like Google and Microsoft do, my research allows me to step back and say ‘hang on, this stuff is really cool but equally, it’s a really big problem. Let’s just take a minute to examine it.’
You mentioned the impact of AI on routine tasks. What specific challenges or problems associated with AI are you most interested in?
There are lots of ways in which AI has the potential to improve our productivity, to the extent that it’s also raised concerns about jobs being displaced in the future. However, whilst AI can generate impressive results, currently over time it might start to lack the same spark that you might see in works produced by a human over their lifetime. I’m interested in understanding how we can capture and integrate that human touch, beyond just generating versions of what is learned.
Another issue with transformer-based AI is that it’s sometimes not very good at getting things 100% right. I’m interested in how we can ensure that models such as ChatGPT can give information that we can trust to be accurate. This is important particularly in areas like healthcare, where AI systems must be reliable if we’re to prevent potentially fatal consequences.
Do you agree with predictions that AI will disrupt our role in the workplace?
Where people are doing routine things, it’s possible that their jobs often could be (and perhaps have already been) replaced by machines. But let’s remember – this is nothing new. At the same time, especially where people use their artistic and creative abilities in their fields, let’s recognise that any disruption from AI could have positive effects on their roles – supporting them, rather than replacing them. What is inevitable is AI being adopted more widely to improve efficiency in the workplace. For example, I am currently working with Kent-based recruitment agency, HR GO, via a Knowledge Transfer Partnership. We’re using AI and machine learning to support people to find suitable employment: both to match candidates more accurately and efficiently to individual roles, as well as to support candidates in finding roles that work for them.
Dr Anna Jordanous is a renowned spokesperson for the field of computational creativity. She has taken part in New Scientist presents, Pint of Science and the chaired the Digital Conversation at the British Library. Her research has been featured in Quartz magazine, the World Economic Forum and Psychology Today. Anna is open to being featured in print, on TV and the radio.