I don't know why but tech industry folks tend to be very resistant to the stupendous complexity of biological systems. The way biology builds functionality into structure can result in incredible space efficiency. Like neurons, for example. People assume the computational aspect of a neuron is only a tiny part of the structure of a neuron, when we have no reason to believe that. That is to say, people think you can equal its ability with a model many orders of magnitude less complex than a neuron. They don't want to think about the implications of that assumption being wrong.

I always find myself coming back to the dragonfly brain. A dragonfly brain needs exactly sixteen neurons to take input from the 30,000 ommatidia in its eyes, use that information to plot the three-dimensional flight path of airborne prey, compute an intercept course, and send those signals to the wing muscles.

How many transistors do we need for that? Input from 30,000 camera pixels, tracking moving objects in 3D space, computing vectors. Now you have a neuron to transistor efficiency ratio. Now multiply that by 86 billion. One brain. AGI's gonna take a hot minute, folks.