An ordinary artificial neuron resembles a chef who strictly follows a recipe: mix the input data in given proportions and output the result. Paul Dirac showed how to turn numbers into objects that break the rule 'the sum doesn't change when you rearrange the addends'. Scientists applied this to the neuron: now the chef has the ability to combine ingredients in any order, even simultaneously — the dish becomes more refined.

Inside a quantum neuron, there isn't a single formula, but a multitude of states existing simultaneously. Using entanglement and temperature tuning of a thermodynamic nature, it trains on seven qubits and makes ten times fewer mistakes than its classical counterpart.

An unexpected ingredient: a mathematical trick replaced the neuron's learning curve with a formula from semiconductor physics, where it described the energy distribution of particles. This allowed the neuron to learn more efficiently. Such neurons will become the basis of quantum neural networks that directly process quantum data. The neuron's operation, including state measurement and quantum information transfer, was tested via numerical simulation before running on a real device. The development involved ideas from Erwin Schrödinger and methods of Seth Lloyd.