The Singularity
The Last Invention Humanity Will Ever Need to Make
There may come a point — perhaps within your lifetime — after which the history of human civilisation becomes, essentially, a preface.
The Idea
The singularity is a borrowed concept. In physics, a singularity is a point where the known rules break down — the centre of a black hole, the first instant after the Big Bang. Mathematician and sci-fi author Vernor Vinge borrowed the term in 1993 to describe something analogous in technological history: the moment when artificial intelligence surpasses human intelligence, after which our models for predicting what comes next simply stop working. The underlying logic is seductive and, depending on your disposition, either thrilling or terrifying. Intelligence is the tool we use to solve problems — including the problem of building better tools. If we ever create a machine that is meaningfully smarter than us, that machine could presumably design an even smarter successor, which designs a smarter one still. The curve bends almost vertically. Ray Kurzweil, who has made a career of tracking exponential trends in computing, predicts this moment arrives around 2045. Others — serious researchers, not fringe figures — think it could be sooner, or that we are already closer than we realise. What makes the singularity genuinely hard to reason about is that it is, by definition, beyond our cognitive horizon. We cannot model post-singularity intelligence any more than a dog can model constitutional law. That is not a rhetorical flourish — it is the core epistemological problem. We are trying to predict the behaviour of something categorically more capable than the instrument we are using to make the prediction.
In the World
In 2017, DeepMind released AlphaGo Zero — a version of its Go-playing AI that was given no human game data whatsoever. Its only instruction was the rules of the game. Within 72 hours, it had surpassed every human player who had ever lived. Within 40 days, it had beaten its predecessor — the version that had already defeated world champion Lee Sedol — by 100 games to zero. This story is often told as a triumph of machine learning. But there is a detail worth sitting with: AlphaGo Zero did not learn to play Go the way humans do. It discovered strategies that human players had never conceived of in thousands of years of play. Commentators watching its matches described some moves as simply baffling — not wrong, just operating from a logic that was not immediately legible. Now scale that up. Go is a closed system with fixed rules. The real world is not. The question the singularity forces us to ask is: what happens when that kind of recursive self-improvement is applied not to a board game but to the problem of making better AI? AlphaGo Zero had a ceiling — the game itself. A sufficiently advanced general intelligence would not. The 72-hour clock, in that scenario, does not stop at superhuman Go. It keeps running.
Why It Matters
Most singularity discourse gets stuck in two unhelpful places: breathless utopianism (AI cures cancer, solves climate change, ends scarcity) or catastrophism (paperclip maximisers, extinction risk, Terminator). Both framings share the same flaw — they assume we will recognise the pivotal moment when it arrives and have time to respond accordingly. The more useful question the singularity poses is not 'will it happen?' but 'how do we make decisions under genuine uncertainty about transformative change?' This is a problem humans are notoriously bad at. We discount futures we cannot vividly imagine. We assume tomorrow will resemble today. Thinking seriously about the singularity — even if you are sceptical — sharpens your instincts for a world that is already changing faster than our institutions can absorb. It asks you to hold two things at once: the humility to admit you cannot predict what comes next, and the responsibility to care about it anyway. That is a useful cognitive posture regardless of whether 2045 turns out to be the year everything changes or just another year.
A Question to Ponder
If a machine became meaningfully more intelligent than you, what would you still be able to offer that it could not replicate — and how confident are you that answer would hold five years later?
Get a new one of these every morning.
Start learning with Thinkable