a blog about the physics of learning machines.
Most discussions of artificial intelligence stop at mathematics and code. I start one level deeper — at the physics. Every machine that learns, whether a living organism, a neural network, or a future quantum computer, is a physical system. It obeys thermodynamics. It pays for every bit it stores and every inference it makes in energy and entropy. That physical price tag, usually ignored, explains a surprising amount: why learning is not the same as memorising, why ideas are not free, why time runs one way, and why intelligence has an energy budget.
The writing here follows three threads:
- Quantum — how quantum systems compute, sense, and carry irreducible uncertainty: photonic qubits, superposition, quantum noise.
- Learning & Thermodynamics — learning as entropy reduction, the thermodynamic cost of computation, and the difference between machine learning and a machine that genuinely learns.
- Consciousness — how self-awareness can emerge, by necessity, in any embodied agent that must model itself.
The aim is straight talk about hard ideas: rigorous enough for physicists, clear enough for the curious. No hype, no hand-waving — just the physics, followed where it leads.
About the author
I am a quantum physicist. My current positions are: University of Sussex and National Quantum Computing Centre.
Fermionik 