This week on the masters we were introduced to the area of computational creativity and AI, topics getting ever more attention as this disruptive technology changes the world around us. I ended up going on an internet exploration on the subject, and it culminated in writing this text to help me organise my thoughts on AI and its possible ramifications.
This is not a ‘typical’ entry in the CRJ, more a researched and critical reflection on the subject of AI and how this affects UXD. Click here to read my take on the challenge and reflection on the week as a whole.
A lot of the fears surrounding AI are that machines will one day be creative like us and replace humans. I first wanted to understand what creativity meant, how it is used and how it could be transferred over to a machine. The Oxford Learner’s dictionary gives the following definition of creativity:
creativity: the use of skill and imagination to produce something new or to produce art.
Skill and imagination. I understand this as the ability to link up concepts in new ways, sometimes it even feels like an idea comes to you out of the blue. Austin Kleon says “All creative works build on what came before. Nothing is truly original.” (2012: 7) He explains in his book Steal like an Artist that if we get enough different sources to inform our creative work, then we will be called original because of this varied mash-up of ideas.
Creativity is also the ability to consider realities that don’t yet exist, ideas over which we then apply our critical thinking to decide if we’ll take them forward. Tanya Krzywinska summed up the interplay between imagination and critical thinking perfectly by saying “In the development process [of a project] creativity and analysis should be regarded as two sides of the same coin”.
This linking of ideas and the ability to predict which resulting ideas will work and which won’t is exactly what AI engineers are trying to get their machines to do. Machine learning, a subdivision within AI, is actively trying to teach machines to think and learn like people (IBM 2023). Computers have a processing power far superior to ours, they can analyse data a lot faster than us and without getting tired, it is tempting to think that by developing AI we are in the process of creating super-humans.
However, what this technology hasn’t been able to crack yet is the ability to draw from a wide range of sources like we can. Whilst they’re getting cleverer, they still have what Jim Guszcza describes as “narrow” intelligence. This means that they’re getting better at learning and thinking yes, but only within the data field that they have access to, they cannot reach out of their own accord and consume new data (Guszcza 2018).
Of course technology is advancing, so I thought I’d try out a bit of a meta test and ask an AI bot directly what it thought about its creative abilities compared to humans. I chose an AI engine that has been getting a lot of attention and clout recently: chatGPT. Below are some of the more illuminating answers:
<aside> 💡 me: is ai better than humans at coming up with ideas? chatGPT: No, AI is not better than humans at coming up with ideas. AI can assist in idea generation by providing information and suggesting possibilities, but it lacks the ability to think creatively and understand context, emotions, and personal experiences that are often the source of new and innovative ideas. Ultimately, human creativity and imagination are unique and cannot be replicated by AI.
</aside>
<aside> 🧠 me: ai can't think critically? chatGPT: AI models, such as myself, are not capable of independent thought or critical thinking in the same sense that humans are. AI models, including language models like me, are trained on large amounts of data and use statistical algorithms to generate responses based on patterns in that data. While I can provide information and assist with research, I cannot form my own opinions, make independent decisions, or think critically in the same way that a human can.
</aside>
Critical thinking involves evaluating information, considering different perspectives, and making informed decisions based on evidence. This requires the ability to understand context, identify biases, and draw conclusions based on a range of factors, something that AI models are not capable of doing on their own. While AI tools can assist with research and information retrieval, it is still up to the individual to engage in critical thinking and independent analysis to make informed decisions.
It is important to note that chatGPT had a knowledge cut-off in 2021 and therefore may not have the most up to date capabilities on the market, but it does suggest that the ability to reflect on consumed knowledge remains in the field of humanity for now. AI bots can try and predict if something will be successful or not based on past patterns through various forms of machine learning and data analysis, but they cannot think and feel like people, so they cannot understand what will and won’t work for people, this is where the role of human-centered design comes in.