At JUXT, any in-person get-together is an excuse for a tech conference, and our Christmas party is no exception. This year’s venue was the Quantum Untangled exhibition at the Science Gallery London, and it inspired me to share some of the unconventional route that brought me into software engineering: filing cabinets, lightbulbs and an arts degree.

Much of the talk draws on the books of David Deutsch, particularly The Beginning of Infinity. Deutsch’s thesis is that the scientific revolution wasn’t just a significant cultural moment for humanity; it was a significant moment in the universe itself. Once humans became focused on identifying good explanations, the world entered a period of potentially unbounded progress.

This framing helped me articulate something I’ve been puzzling over: the sharp polarisation between those who wield AI despite its flaws and those who reject it despite its extraordinary successes. I wonder if the divide has to do with intuitions about the quality of explanations. Human progress is full of incredibly useful models that turned out to be wrong. The geocentric model of the universe, for example, was a sophisticated, predictive system that served astronomers well for centuries. Seasoned technical people, especially, have a deep aesthetic objection to anything that seems predicated on flawed explanations.

In the talk I share some of my own disappointments and concerns about the current state of the AI industry, but above all my excitement that we simply can’t know where the next stepping stones will come from. Whether current AI models turn out to be an evolutionary dead-end or the basis of ever more powerful intelligence, we can only connect the dots in hindsight.

The technology for making reading glasses led to Galileo’s telescopes and the end of the geocentric model. The valves for amplifying telephone signals enabled the first electronic computers. We didn’t know, and we couldn’t have known.

We’re still early, and there will always be problems to solve. As Alan Kay said, the best way to predict the future is to invent it.