They can't. Biological neural networks have no resemblance to the artificial neural networks of the kind used in LLMs. The only similarity is based on a vague computational abstraction of the very primitive understanding of how brains and nerve cells worked we had in the 50s when the first "neural network" was invented.
Real brains have a very different structure and composition. Neurons aren't monotonic, there are loops, etc.
They can't. Biological neural networks have no resemblance to the artificial neural networks of the kind used in LLMs. The only similarity is based on a vague computational abstraction of the very primitive understanding of how brains and nerve cells worked we had in the 50s when the first "neural network" was invented.
These pieces of "understanding" mostly don't even applied to other LLMs.