On the arm of Swiss tennis player Stan Wawrinka is tattooed a quote by Samuel Beckett: “Ever tried. Ever failed. No matter. Try again. Fail again. Fail better.”
This excerpt from novella Worstward Ho seems motivational and suggests that perseverance is needed for success. However, the word failure carries a weight with it, especially if used as a label, as if it were an essential part of someone.
Yet, in evolution, the creative arts, engineering and education, failure is a process – without which success is not achievable.
“Error” might actually be a better term than failure, because error generates variation. And this variation is important in understanding the uniqueness of human creativity.
Generative AI can create fashion models, award-winning art and actors. But generative AI lacks the artist’s drive, their ability to reflect and know the significance about why and for whom the art is being created.
If we consider creativity as a process, then in order to create new and novel art, errors, mistakes, dead ends are required. In short, failure.
Generative AI also cannot understanding concepts such as aesthetic failure (when musicians use failure as a catalyst for improvisation), or have the desire to connect with an audience in a live performance. Creation can be outsourced but human creativity and the impulse to connect cannot.
Perfectionism is an illusion
Learning from mistakes in not a new idea in teaching, but with the rise of generative AI the temptation for both students and educators might to see generative AI as a way to eradicate failure, a guarantee of high grades at school and university.
However, this risks not providing students with the experiences they will need to be lifelong learners. British psychiatrist and cyberneticist W. Ross Ashby wrote: “The whole function of the brain is summed up in: error correction.”
Here, the key to understanding the brain is not in the error, but the process of correcting the error. Similarly, in his book To Engineer Is Human: The Role of Failure in Successful Design, the engineer Henry Petroski argues that failure is vital to the advancement of engineering and design, because it drives process.
Not that anyone deliberately designs bridges to collapse, but the knowledge of how to put things right comes from understanding why things went wrong. Petroski also argues that prolonged success leads to failure, but this is because of complacency.
In deciding what we want from AI, complacency (not failure) is our biggest enemy. Across many domains failure is not just necessary, but vital for success.
For example, a research study has found that both AI models and human dermatologists perform worse on images of dark skin tones and uncommon diseases when presented with a set of diverse skin images. This highlights the problems of a lack of exposure to variations in skin types and rare skin diseases in both AI trained datasets and humans.
Driverless vehicles have issues with merging into traffic and halting because they do not have a mental representation of the intentions of other road users.
By contrast, humans understand driving as a social, interactional and transactional endeavour – as much as a technical one – and, so, find ways to negotiate, to yield and say thank you.
Appreciating this a powerful counter narrative to perfectionism in all its guises. The most seductive of which is perhaps the promise of an AI-created utopia.
The question is whose vision of paradise is this and what are we forsaking by not questioning it. What we do risk losing by not striving, by not making (or accepting mistakes), of seeing beauty in imperfection?
The fallacy is that we have no agency, that technology cannot be imbued with moral ambition. However, history shows us that humans can and do shape technologies. For example, the printing press was repurposed from publishing books to printing newspapers – thereby creating the means and a mechanism for a free press.
So, there is no such thing as technological inevitability. We can decide what the relationship between humans and AI will look like – through consumer choice, the ballot box and legislation – and with it all the groundbreaking, creative and beautiful mistakes it will bring.
The post “Why failure is a necessary ingredient for success – especially in the era of AI” by Thusha Rajendran, Professor of Psychology, The National Robotarium, Heriot-Watt University was published on 01/22/2026 by theconversation.com