Is it like a brain?

The other question that has been running underneath, and it deserves a final boundary rather than a shrug.

Where the comparison is real. One property genuinely holds, and you met it with the delete test. Both a trained model and a human brain hold information as patterns spread across enormous numbers of connections, rather than in labelled locations. Neither has a row you can delete. Both degrade rather than losing specific files, and both can recover a whole from a fragment. That is not a loose analogy; it is the same organising principle.

The historical debt is real too. The artificial neuron was inspired by a sketch of the biological one, with inputs, weights, a threshold, and an output.

Where it stops. Almost immediately after that.

A biological neuron is a living cell, communicating through chemistry as well as electricity, with thousands of connections that physically grow and retract. An artificial one multiplies numbers and adds them. The resemblance is at the level of a stick figure to a person.

A brain learns continuously, from a handful of examples, while running the body it lives in. A model is trained once and then frozen, identical between conversations. Backpropagation, the mechanism at the heart of all of this, is not thought to be how brains learn.

A brain runs on roughly the power of a dim light bulb. Training a frontier model consumes the electricity of a small town for months.

And the direction of the debt is often reversed in popular accounts. Neuroscience did not hand over a working design. A very simplified idea from the 1940s was borrowed, and what happened afterwards was engineering, driven by what could actually be trained.

What the comparison comes to. One shared principle about how information is held. Everything else, components, learning, embodiment, energy, development, profoundly different.

So: inspired by a brain, and not a model of one. The next time you see a headline about a digital brain, you know exactly which one property it is stretching.