The Centaur Advantage: How the AI-Human partnership could change us for better and worse
AI is making us smarter and more capable, but at what cost? A compelling look at the fascinating history and future of human-AI collaboration.
The rise of increasingly capable AI promises to transform how we do everything. AI is being woven into the fabric of our lives through smart assistants, copilots, chatbots, and more. The rise of frameworks like Langchain and boxcars.ai promises a future where all apps will include AI. While this could augment our abilities greatly, it also poses risks to skills and ways of thinking we’ve long relied upon. How will this human-AI symbiosis shape us as individuals, employees, and a society? Will specific skills be rendered obsolete, or will we develop new human talents to complement AI?
In 2018, technology pundit Ben Evans proposed that our mental model for an AI assistant should be like getting a summer intern to help us. At the time, he believed AI interns had the capacity of a 10-year-old, capable of handling some tasks but requiring close supervision to avoid going off the rails. Fast forward a few years, and these AI interns have made remarkable strides, passing advanced placement tests and excelling in a wide range of tasks. As our reliance on AI assistants grows, how will this symbiotic relationship shape us positively and negatively?
Thanks for reading BoxCars AI! Subscribe for free to receive new posts and support my work.
To explore this question, let’s delve into the fascinating history of AI-human collaboration, shedding light on our evolving partnership.
The birth of Centaur Chess
On May 11, 1997, the world was captivated by the defeat of chess grandmaster Garry Kasparov at the hands of IBM’s Deep Blue, a chess-playing computer. This watershed moment heralded a new era in which AI could outperform even the most skilled human players. However, Kasparov wondered about the potential of combining human intellect with AI, rather than pitting them against each other. In 1998, he organized the world’s first Centaur Chess tournament, inspired by Greek mythology’s mythical half-human, half-horse creatures. The contest welcomed players of all skill levels, from experts to novices, forming teams of humans and AI engines to compete against one another. At first, top-rated players dominated by using AI “engines” to battle opposing teams of grandmasters and engines.
But, in 2005, an unexpected event caught the chess world’s attention. Two amateurs, Steven Crampton, and Zackary Stephen, teamed up with three computers to form a group called “Zacks”, entering the Freestyle Chess tournament. As a dark horse, Zacks astonished spectators by defeating teams that included grandmasters and some of the best AI chess engines, ultimately winning the competition.
The victory of Zacks demonstrated that machine-assisted success wasn’t limited to grandmasters wielding AI engines; amateurs who knew how to collaborate effectively with AI could also achieve remarkable results. This revelation has far-reaching implications for other areas of human endeavor, such as writing. While a spellchecker doesn’t inherently make someone a better writer, a mediocre writer who can skillfully prompt an AI assistant might produce superior content compared to a more talented writer who cannot. (In case you are wondering: yes, this essay is powered by performance-enhancing AI assistants) In essence, the nature of human intelligence needed transforms when we work alongside machines.
Indeed, what we do changes us.
London taxi cab drivers
Take, for example, the arduous process of becoming a licensed taxi driver in London. Aspiring drivers spend 3–4 years navigating a labyrinth of 25,000 streets, committing them to memory to pass the notoriously difficult “The Knowledge” test, which has a mere 50% success rate. Inside the brains of these drivers, the hippocampus, a part responsible for spatial intelligence, plays a crucial role in their ability.
In 2000, researchers Eleanor Maguire and Katherine Woollett at University College London investigated whether becoming a London taxi driver would physically alter the drivers’ brains. They compared two groups: one consisted of aspiring drivers, while the other was a control group of non-drivers of the same age. At the beginning of the study, both groups had similarly sized hippocampi. Four years later, the successful taxi drivers’ hippocampi had significantly increased in size, demonstrating that their brains had adapted to the skills they needed to acquire.
However, the advent of GPS may have made spatial intelligence less critical in the intervening years. In a study conducted 20 years later by Louisa Dahmani at Harvard Medical School in the US, it was discovered that individuals who relied on GPS for navigation had poorer spatial memory, landmark encoding, and cognitive mapping abilities. Over a three-year follow-up period, those who continued to use GPS exhibited an even steeper decline in spatial memory compared to non-users.
But as you shed some skills, it creates room for new and possibly ‘better’ skills.
The Mechanical Loom
The invention of the mechanical loom in the late 1700s offers inspiration for navigating the rise of advanced AI. The mechanical loom automated the intricate handweaving techniques that craftspeople had relied upon for millennia. This eliminated the need for many physical skills involved in handweaving, such as manually passing weft threads through warp threads. However, the mechanical loom also enabled new creative skills, such as designing punch card patterns for Joseph Marie Jacquard’s programmable power loom, introduced in 1801. These punch cards allowed weavers to create highly complex woven patterns.
So within a few decades of the mechanical loom’s introduction, it both eliminated some longstanding skills of handweaving, as well as introduced new creative skills. This example from the Industrial Revolution highlights how as technology progresses, both losses and gains occur for human skills.
These tradeoffs are part of our journey
Abundance of books makes men less studious; it destroys memory and enfeebles the mind by relieving it of too much work.
Hieronimo Squarciafico, Memory and Books, 1477
Remember those taxi drivers from a few paragraphs above? Their intense practice and memorization enlarged the rear of the hippocampus, boosting spatial memory. But this neural remodeling came at a cost: drivers struggled with visual memory, worse than others at recalling a complex doodle. It seems the brain, like a muscle, can grow stronger in certain ways through rigorous exercise, but not all: gains in some areas may mean tradeoffs in others.
As we ponder the implications of these findings, we must ask ourselves: What kind of future awaits us in a world where AI assistants increasingly supplement our own capabilities? Will we see a decline in certain cognitive skills as we offload more tasks to our AI counterparts? Or will this partnership usher in a new era of unprecedented human achievement, as we harness the power of AI to augment our own intelligence?
The era of increasingly capable AI is unwritten, within our power to craft for a future of human flourishing, augmented by AI as a tool igniting human talents, not overriding them.
More content at PlainEnglish.io.
Sign up for our free weekly newsletter. Follow us on Twitter, LinkedIn, YouTube, and Discord.
Interested in scaling your software startup? Check out Circuit.
