Human intelligence is not computable
Mark Buchanan
DOI: https://doi.org/10.1038/s41567-024-02533-w
IF: 19.684
2024-06-15
Nature Physics
Abstract:To define his dynamical system, Penrose imagined a phase space of states defined by the different sets of polygons. The system moves between these states using two different rules, one rule holding if the answer to the tiling problem for the current state is 'yes', and another holding if the answer is 'no'. This particular definition isn't important. The point of Penrose's construction was to demonstrate a possibility: that a dynamical system can be fully deterministic, yet non-computable. Hence, no digital computer can simulate such a process. Deterministic is not, as is often believed, the same thing as computable. This is, I have always thought, a profound — and somewhat underappreciated — point. After all, as Penrose examined in his book, the laws of physics as we currently know them seem to be computable; we can always simulate known systems with arbitrary accuracy on a digital computer, even if doing so might be impractically slow or expensive. Yet the world doesn't have to be this way, and, in future, we may well discover some real-world processes that really are beyond computation — things we cannot simulate with computers of the current kind (that is, those based on the notion of a Turing machine).
physics, multidisciplinary