Note ·

Information is the fundamental unit

PhilosophyIntelligenceNeuroscienceResearch labsMachine learning

The idea that organises Hassabis, in Mallaby’s telling, arrived in his third year at Cambridge, in tutorials with John Daugman that he still calls nirvana sessions. Through Shannon’s theory of information, information as the opposite of uncertainty, something you can quantify, store and move, he came to the view that information, not matter or energy, is the fundamental unit of reality, and that a theory of everything would therefore look less like an equation and more like a program. Hofstadter had primed him for it: what matters about a brain is not the stuff but the patterns in the stuff.

The second half is the one that became a company. If information is the bottom layer and computation the layer above it, and if the computation humans could design had stalled on the problem of induction, learning from messy examples rather than deducing from clean rules, then the answer was a third layer: machines that work out their own computation. Programs that design programs, a decade and a half before anyone would fund one. His doctorate followed the same thread into the brain, where he found memory to be reconstruction rather than playback, and concluded, with Kant, that the mind builds reality more than it records it.

I do not know whether information is the fundamental unit, and I notice that the people most certain of it are the ones who build information machines for a living. But it is the first page on this site for the Philosophy topic, and the right one: a metaphysics you can test by building something, which is more than most metaphysics offers.

References

  • Sebastian Mallaby, The Infinity Machine (Penguin Press, 2026), chapters 2 and 3.
  • Douglas Hofstadter, Gödel, Escher, Bach (Basic Books, 1979), the book Molyneux gave him at seventeen.

    Open the full search →