Recent developments in content-generating technologies, such as ChatGPT and DALL-E-2, have raised questions about the creativity and originality of generative AI and the difference with human-level creativity.
Some scholars argue that Generative AIs have reached human-level creativity (Haase and Hanel, 2023), following the psychological definition by Plucker that a creative product is "both novel and useful as defined within a social context.” (Plucker et al, 2004) However, there is debate about so-called weak artificial creativity versus strong artificial creativity. Weak artificial creativity means AI models only imitate creativity. Where strong artificial creativity points to AI models which are genuinely creative.
Looking at originality and spontaneity (Kronfeldner, 2009), rather than solely novelty and usefulness, the view that generative AI has reached human-level creativity falls short. Maria Kronfelder (2009) states that something original does not mean it needs to be new. For example, Newton invented calculus independently from anyone else, even though Leibniz already invented it. Hence, Newton’s discovery wasn’t new, but it was still original since he came up with it on his own without knowing that Leibniz invented calculus before him during that time. Originality is an essential ingredient for creativity. Concerning artificial creativity, this raises the question: is generative AI original or does it only make novel things based on the data it’s been trained on?
The following example shows how conditioning can lead us to be less original. In ideation workshops, I often ask participants to draw a flying horse to get in the right mindset for generating creative ideas. They almost always draw a Pegasus. When asking DALL-E-2 it also generated Pegasus images. Both people and AI are conditioned by the previous information they consume. However, when asking children to draw a flying horse, they draw a horse with rocket launchers for feet or a horse with balloons attached to it. The children’s answers seem intuitively more creative. This is because they have not been influenced by previous conditioning and can still make original associations, although the idea of a horse with rocket launchers for feet isn’t something new (it’s probably been drawn before).
I find it immensely important that we still have a sense of what it is to be human, as creative beings, and gain more clarity on the role of generative AI in society. Only then can we find the right way of relating to these technologies. With generative AI evolving at a rapid pace we have to reconceptualize what creativity means, also looking at what it is to be creative rather than to behave creatively. To understand what it is to truly be creative as human beings, we can shine a light on our relationship with generative AI, which seems to only behave creatively.
I am deeply inspired by Taoism, which holds a radically different perspective from our Western way of thinking about creativity. The latter often sees creativity in terms of a mechanistic process which can be reduced to certain conditions. A Taoist vision of creativity, developed by David L. Hall (1978), sees the characteristics of creativity as freedom and reflexivity. Hence, creativity is something which arises in an effortless state or without deemed action. Chang Chung-yuan finds that “the creative process of the universe is also the creative process of the poet, who has transformed his ego into self and thus has become part of the universe.” (Chang, 2011, p.200) Indeed, from a Taoist perspective, the creative and spontaneous activity of Tao, often translated as the way or course of nature, mutually arises with human creativity. Similarly, we find this view in the works of Plato (4th century BCE) where Socrates observed great poetry as arriving through divine inspiration. Creativity can be regarded as being interrelated with the concept of self-realisation, to go beyond the ego and transform as human beings into a genuine self. The goal of self-realisation is “to be free from the confusion of external conditions” (Chang, 2011, p.93).
I propose we all start to look at how conditioning, both in training generative AI models and learning in humans, is related to different types of creativity. To become truly creative and be divinely inspired, we must go beyond the individual ego and mechanistic models of creativity. I wholeheartedly think generative AI models will never truly be creative, but they will remain tools that can behave as if they are.
DISCLAIMER: This blog post was written without generative AI technologies.
References (in order of appearance)
Haase, J. and Hanel, P.H., 2023. Artificial Muses: Generative Artificial Intelligence Chatbots Have Risen to Human-Level Creativity. arXiv preprint arXiv:2303.12003.
Plucker, J.A., Beghetto, R.A. and Dow, G.T., 2004. Why isn't creativity more important to educational psychologists? Potentials, pitfalls, and future directions in creativity research. Educational psychologist, 39(2), pp.83-96.
Kronfeldner, Maria E., 2009, “Creativity Naturalized”, The Philosophical Quarterly, 59(237): 577–592. doi:10.1111/j.1467-9213.2009.637.x
Hall, D.L., 1978. Process and anarchy: A Taoist vision of creativity. Philosophy East and West, 28(3), pp.271-285.
Chang, C.Y., 2011. Creativity and Taoism: A study of Chinese philosophy, art and poetry. Singing Dragon.
Cooper, J.M. and Hutchinson, D.S. eds., 1997. Plato: complete works. Hackett Publishing.
Comentários