The Wisdom of the AI Crowds

Professor Geraint Rees speaking ahead of Scientific Synergy: Unlocking the UK’s potential to become a scientific industrial powerhouse. UCL Policy Lab in partnership with Onward. 

Predictions are difficult, especially about the future. When we try to make predictions about who will win the Premier League, Strictly Come Dancing, or the General Election it turns out that collecting the overall opinion of many individuals is often a better predictor than consulting a single expert. The phrase “the wisdom of the crowds” has come to describe this phenomenon. But in a world where artificial intelligence models can ingest vast quantities of data and make increasingly accurate predictions, the wisdom of the crowds seems increasingly outmoded. Why consult a human expert when ChatGPT can answer your question?

But the original meaning of the wisdom of the crowds was rather different. The ancient philosopher Aristotle used it to refer not to predicting the future, but to debating the present. He suggested that bringing people together to debate and discuss ideas would lead to better collective wisdom than relying on individuals. Today, as party conference season is in full swing, we see this principle in action as ideas and philosophies are vigorously debated and discussed in conference halls across the UK.

And it is no coincidence that political parties gather their supporters for these conferences in our great cities. Cities are one of the ways in which people come together in huge and diverse groups to discuss, debate, argue and disagree on many matters.

Our universities are also places which seek to encourage healthy debate. My own university was founded on the principle that admitting diverse people from different backgrounds was a way to live well with difference. This was a radical principle in 1826, where other universities encouraged conformity of thought. Today in the 21st century, we are still actively encouraging our staff and students to ‘disagree well’. We believe the need to disagree well is a core skill for our students to learn as citizens of democratic societies. They need to be able to listen carefully to those you disagree with, face up to disagreement, and learn from the process. Disagreeing well is a core skill in modern liberal democracies, neither promoting conformity nor preventing consensus. Rather, it celebrates diverse viewpoints and encourages useful knowledge.

In such a world, what has artificial intelligence to contribute? Perhaps less than we might think. The number of ‘artificial intelligence’ language models in the world can be counted on our fingers, so are hardly a crowd. Their contributions may be accurate (when they are not hallucinating their answers) but they are hardly diverse. So, as we debate and discuss the future of our societies, we might like to reflect upon how artificial intelligence might be developed so we think of future AI complementing the true wisdom of crowds, not replacing them.