>>936525586
First you should understand different neural networks mirror different parts of the brain, with the ambition of "sentient" ai merely being an organized network of these systems. E.g. SD is the optical cortex, with its UI serving as cortical feedback neurons and its image processing being akin to the Drosophila visual circuit. Llms are portions of the frontal cortex, but most lack the CA1 and CA3 regions needed for long term association and memory. It ought to intrigue everyone that we continue to use such delaminated systems when fully organized systems have been in development since the 1970s.
Carrying this metaphor forward, the neural network model simply doesn't account for the peripheral CNS like the vagus nerves and autonomic nervous system, and the state of biophysics indicates the only way to achieve this kind of parity is by using wetware. So, even in the reductionist framework of the brain as a computer, a computer we make will never be able to gather subtleties about the environment and emotional reactions like we do even with logical routines that emulate this information.
However, the more important point you have encountered is that there is 0 evidence that human brains are limited to electromagnetic and thermal laws that restrict traditional computing. Neurons are structures that contain heterogenous nanoscale structures where subatomic forces like casimir effects (to name 1 example) dominate. Physics indicates the brain is much more complex than the so-called scientific revolution has described and draws upon complex subatomic, subliminal and superliminal regimes that allow information processing, sotrage and transmission beyond anything a solid state computer is capable of today. This can change, and we have been warned of the consequences of machines acquiring parity with human consciousness-- something we in modernity see in a restricted and flawed framework.