Neural drift keeps asking what a stable brain code really means
If neural representations keep shifting, intelligence may be less like a static codebook and more like a living control process.
Neural drift is one of those brain-science ideas that quietly changes the metaphor. Instead of assuming that stable behaviour requires stable individual-neuron encodings, the evidence suggests that populations can move while the organism still remembers, recognises, and acts.
That matters for brain-computer interfaces because decoders trained on today’s neural activity may decay tomorrow. It also matters for AI because representation drift is usually treated as a training problem, while biology may treat it as a feature of robust adaptive systems.
The most interesting bridge is between neuroscience and engineering: how do you design systems whose internal representations can change without breaking the external skill?