Why Computers Aren’t Writing Shakespeare Yet: Reflections on Margaret Boden’s AGI12 Talk
Back at the 2012 Oxford Winter Intelligence conference, cognitive scientist Margaret Boden gave a talk on creativity and Artificial General Intelligence (AGI).
If you’ve been caught up in the modern race to build autonomous systems that can code, paint, or write poetry, Boden’s framing feels remarkably grounded – and quite humbling.
Creativity Isn’t Magic
Boden starts by stripping away the romantic myth that creativity is some mystical spark reserved for a tortured genius. True, history remembers the outliers, but ordinary creativity is a baseline part of being human. We all do it – whether we’re improvising a joke, making an off-the-cuff analogy, or rearranging our living rooms.
Her definition of creativity is delightfully straightforward: it’s the ability to come up with ideas that are new, surprising, and valuable.
Maggie thought that if an AGI doesn’t have creativity, it doesn’t have general intelligence, one can’t separate the two and actually building that capability into a machine requires understanding that creativity isn’t just one thing – it comes in three very different flavours.
The Three Flavors of Creativity
Combinational Creativity: This is what most people think of – shuffling familiar ideas into unfamiliar combinations, like poetry or collages. It’s easy to get a computer to randomise data, but getting it to understand relevance and value? That’s an entirely different beast. Boden points to Shakespeare describing sleep as a knitter repairing a ravelled sleeve of care. A computer can parse text, but capturing that kind of poignant, human resonance is a monumental wall to climb.
Exploratory Creativity: This is where about 95% of professional art and science happens. You take an existing style or rule set (like Impressionist painting or a branch of mathematics) and explore its outer boundaries. This is the “easiest” for AI to tackle because if you can explicitly program the rules of the style, a computer can happily map out that space.
Transformational Creativity: This is when you change, delete, or negate the rules themselves – like Picasso inventing Cubism or Kekulé realising the benzene ring is a closed loop. People often assume computers can’t do this because they just follow code. Boden pushes back here, noting that evolutionary algorithms and genetic programming can produce radical mutations. For Maggie the bottlenecks weren’t generating random transformations – it was figuring out which ones are valuable and worth sustaining.
Why Boden Was Skeptical of Fast Timelines
Listening to her push back against the ultra-fast AGI timelines floating around (like Ray Kurzweil’s mid-century predictions) is refreshing. While tech optimists (myself included) lean heavily on exponential-esque hardware growth and possibly algorithmic breakthroughs (which are hard to predict), Boden argues that we’re bottlenecked by deep conceptual problems.
She takes a subtle swipe at neuroscience along the way, calling a lot of brain-scanning “natural history” or “theoretical fishing expeditions” – fun to look at, but lacking the deep psychological and computational theories needed to actually explain how the mind works.
AI Joining the Moral Community
What really sticks with you at the end of the talk is her final thought during the Q&A.
If we ever do build something beyond tool-AI that achieves true human-level intelligence and creativity, we’re opening up a massive philosophical can of worms. Maggie suggested that to genuinely accept an AI as creative and intelligent on our level means we’d eventually have to accept it into our moral community – acknowledging its rights, responsibilities, and perhaps even valuing its interests alongside our own.
Until then? As Boden playfully warned her audience when talking about solving these deep computational hurdles: don’t hold your breath.
See the interview I did with Margaret Boden at the conference: https://www.scifuture.org/margaret-boden-creativity-and-artificial-intelligence-interview/
