Artificial intelligence is making continued and complex strides, impressing even its most knowledgeable critics
As a mathematician, Marcus du Sautoy has always considered creativity to be a protective tool against the notion that machines can’t do his job because mathematics is not just about computations; it’s actually much more creative. He used to believe that computers couldn’t do his job due to this creativity aspect.
However, a few years ago, he saw something that began to worry him – what he believes was the first sign of creativity emerging from a machine. This occurred in the context of a game when a piece of AI challenged the world’s best human player at the game of Go. Go is a game played on a 19×19 grid, where players place black and white stones and try to surround more territory than their opponents. You may recall when Garry Kasparov was beaten by Deep Blue at chess in the 1990s. Go is far more complicated to write code for and computer scientists traditionally considered it nearly impossible to do so. If you ask a Go player why they’re placing a stone in a particular spot, they find it hard to articulate a reason. They rely on intuition and pattern recognition, making it difficult to write code that can predict or counter moves. What changed? A few years ago, the style of writing code shifted from a top-down approach, where humans wrote specific instructions, to a bottom-up method, where the code can learn, change and evolve. This is called machine learning or deep learning, and it’s how AlphaGo learnt to play Go. AlphaGo studied human games, understood critical moves and eventually started playing itself. Gradually, a piece of code emerged that could beat all other versions, and DeepMind, the developer, believed it was good enough to challenge the world’s best player, Lee Sedol. Sedol was dismissive of Alpha Go initially, having never seen any code that could play Go well. He was surprised to lose the match 4-1, with the one game he did win being the most valuable victory of his career. We’ve become accustomed to computers beating us at games, but the most remarkable aspect of this match occurred during the second game. Lee Sedol, playing white, made a move and took a break. AlphaGo,playing black, placed a stone five rows in from the edge – a move that surprised and puzzled the commentators and Sedol.
Learning on the go
Du Sautoy watched these games obsessively on YouTube Live because he saw similarities between Go and mathematics, both involving pattern searching. He thought that if a computer could play Go at a high level, he might be next in line. Margaret Boden, a renowned scholar, defines creativity as something that is new, surprising and valuable. In the context of AlphaGo, the novelty of the move was clear, the surprise was evident among commentators and players, and the value was demonstrated when the move helped AlphaGo win the second game. Some might argue that the creativity lies with the humans who wrote the code, but Du Sautoy disagrees. If a human had seen the line of code responsible for the unconventional move, they would have likely deleted it, believing it to be a mistake. The move emerged from the code’s learning process, so he believes we should credit the code’s creativity, not the humans who initiated the process. This development is exciting because it shows that AI could help us discover new and exciting ways to approach tasks, like AlphaGo showing us a superior way to play Go. There have been three revolutionary moments in Go’s history: one in the 16th Century, one in the 20th Century, and now AlphaGo, which has taught us a new way to play the game at a high level. The exciting potential of AI is to show humans how to do things in new and innovative ways. Once Du Sautoy saw this story of AlphaGo making what he considers the first creative move by a computer, it set him off on a journey to explore the potential of AI in our own creativity.
Strength in collaboration
The result is Du Sautoy’s book The Creativity Code: Art and Innovation in the Age of AI. Published in 2019, it delves into the intersection of art, creativity and artificial intelligence, exploring the ways in which AI is transforming our understanding of creativity and the potential of machines to produce original, innovative work in various fields, such as art, music, science and mathematics. “The implications of AI’s creative abilities extend beyond just games,” he says. “This breakthrough could pave the way for advancements in numerous fields. By embracing AI’s potential to enhance human creativity, we can unlock new possibilities and push the boundaries of what we thought was achievable.” In the book, Du Sautoy examines the development of AI and its increasing capability to learn, adapt and create. Besides discussing groundbreaking AI achievements that have come a long way since its publication, he also addresses the ethical and philosophical questions surrounding AI and creativity, including whether AI can truly be considered creative and the implications of its advancements for human creativity and innovation. Ultimately, he believes that the collaboration between humans and AI could lead to the development of groundbreaking ideas and solutions that might have otherwise remained undiscovered. So instead of viewing AI as a threat to our creativity, we should see it as an opportunity to expand our own creative horizons. Indeed, the emergence of creativity in AI, as demonstrated by AlphaGo, has profound implications for the future of human innovation. By recognising and harnessing the potential of AI, we can open new doors for creative collaboration and exploration, ultimately enriching our understanding of the world and our place in it.
Text | Eugene Yiga Photography | maxuser
For more information, go to eugeneyiga.com.