The 3 levels of creativity in the AI Era | Oxford, Marcus du Sautoy
In a Nutshell
Marcus du Sautoy defines three levels of creativity—exploratory (pushing existing rules), combinational (merging domains), and transformational (breaking rules to create new frameworks)—and argues AI excels at the first two but struggles with the third. The AlphaGo Move 37 exemplifies how AI can produce transformational creativity by discovering strategies humans would have dismissed, fundamentally changing Go play. Ultimately, AI functions as "augmented intelligence" rather than a competitor, best serving as a collaborator where humans introduce efficiency shortcuts to complement the machine's exhaustive processing.
These notes were generated by AI and may contain inaccuracies.
Marcus du Sautoy is a professor of mathematics at the University of Oxford and the Simony professor for the public understanding of science. His role serves as a bridge between academia where developments occur and society which will be impacted by new technologies, particularly artificial intelligence.
AI has been remarkably successful, prompting questions about whether there is anything it cannot do. Creativity is frequently cited as something AI could never achieve, with many viewing creativity as a unique expression of human identity. However, the concept of creativity is difficult to pin down, making it challenging to determine whether AI can be creative.
There are three distinct forms of creativity. Exploratory creativity involves taking existing rules and pushing them to their limits within that rule set. Combinational creativity involves finding new things by combining different areas or disciplines. Transformational creativity is the most difficult and rarest form, where conventions of the past are broken and something emerges seemingly out of nowhere.
Humans are described as a lazy species, comparable to a lion that sits around all day and then does a short burst to capture prey. This mentality often leads to incredible innovation when facing problems, as people seek clever shortcuts rather than laborious approaches.
Du Sautoy did not immediately fall in love with mathematics in school due to its focus on technical aspects like multiplication tables. At age 12 or 13, a teacher showed him the beauty and creative side of mathematics, revealing how mathematics underlies everything, especially nature.
The Fibonacci numbers (1, 1, 2, 3, 5, 8, 13) demonstrate how simple patterns explain nature's growth. Mathematics differs from other sciences because it allows exploration of worlds without physical reality, whereas other sciences must match observed reality.
Ancient Greeks developed Euclidean geometry, which is flat. In the 19th century, mathematicians created new geometries where triangles behave differently: on a sphere, triangles add up to more than 180 degrees, and hyperbolic geometry (described as saddle or Pringle crisp geometry) allows triangles to do different things. These are stories about universes with no physical reality.
Mathematics shares characteristics with novelists or science fiction writers who establish rules and explore their consequences. Du Sautoy fell in love with mathematics because of its creative side.
The word creativity is hard to pin down, presenting the first challenge in determining whether AI can be creative. A clear definition of creativity is needed before asking whether AI can achieve it.
Bach exemplified exploratory creativity by working within the Baroque musical rule set while pushing the musical style to its absolute limits. AI is expected to be very good at this form of creativity.
Du Sautoy uses combinational creativity in his work by attending geometry seminars despite being a number theorist, applying analytical structures from one area to another. A simple example is fusion cooking, combining Asian ingredients with European cooking methods.
Transformational creativity breaks conventions of the past. At the beginning of the 20th century, serialism in music threw away harmonic structure and introduced the 12-tone row. This form of creativity is the rarest and most challenging for artificial intelligence because AI learns from past styles and develops those existing patterns.
AI has helped solve mathematical challenges open for decades. In one case, AI found a counterexample to a conjecture rather than proving it true. AI functions like a digital telescope, allowing visibility into the digital world and revealing patterns. Du Sautoy suggests translating artificial intelligence as "augmented intelligence" rather than artificial intelligence.
Move 37 of game two in AlphaGo's match against Lee Sedol is regarded as the first genuine sign of machine creativity. The move was unconventional, played very deep into the board early in the game, contrary to traditional Go strategy. Commentators initially considered it a mistake, but it proved to be the winning move and has genuinely changed how humans play Go.
The move represents transformational creativity because it broke established conventions about good Go play. AlphaGo revealed that what humans considered the peak of Go strategy was merely a local maximum, and there were higher peaks if different approaches were taken. This strategy emerged from the AI's learning process, not from human coding, and would likely have been deleted by humans who would have considered it a bad direction.
AI is a statistical model generating text based on probability from its learning process. It can show that something happens but not explain why. AI will not write a novel because it wants to express what it's like to be an AI. The intention to express itself would indicate a ghost in the machine.
Humans excel at finding shortcuts when faced with problems. The book "The Art of the Shortcut" explores how this lazy mentality leads to innovation. Mathematics developed from this shortcut mentality.
Carl Friedrich Gauss, at age 8, was asked to add numbers from 1 to 100. Instead of adding sequentially, he paired numbers (1+100, 2+99, 3+98) each equaling 101, recognizing 50 pairs, yielding 5,050. This strategy works regardless of the upper limit, creating an algorithm that efficiently solves the problem.
AI does not mind doing things the laborious way, while humans run out of energy. The ideal combination involves humans introducing shortcuts into AI programs, creating efficient collaboration between human and machine. The central message is that artificial intelligence is not a competitor but a collaborator.
Keep EO in your library
Save the videos and channels worth coming back to, and find them again in one place.
