The interesting guts of machine learning research by emulating the human brain is deeper than we thought. Down at the level of graph theory between your 85 billion Neurons, the Neuron is absolutely doing "dot product of weights and observations", linear algebra, geometry, probabilities, derivatives and partial derivatives of multiple variables with respect to one, and doing gradient ascent, however this perspective is incomplete.
Additional complexity is inside the Dendrite bulb at the molecular level, there is a city's worth of activity and millions of read/write heads in your dendrites, powered by ATP Synthase Turbines, powered by sugar, doing protein folding, rigid carbon chain scaffolding are created as needed for data-transport. Voltage gated ion channels and switches on those molecular electron-powered turbines fueling activity all record your experiences of existing on the arrangement of individual protein molecules.
Your brain's short term and long-term data-storage hard-drive and read/write head resides in the Dendrite bulb, there are millions of them in each neuron. The analog signals passing up and down through the neuron get inscribed on the protein molecules themselves. You've got a little-man in your head with his pen and paper taking notes, but instead of ink and paper, he's using a molecular Nutcracker, and very large molecule neurotransmitters shaped like a fishnet stocking filled with hammers, with ropes attached for crawling around.
By researching protein folding, Brownian motion of molecules at the atomic level, and doing whole-cell computer simulation of the Neuron's dendrite, at the atomic level, there we will uncover more delicious biological secrets of how your brain is able to perform magical feats of complexity that defy explanation, that is the engineering diagram for how discoveries from my brain are flowing directly into your brain via a series of glyphs. It all boils down to mathematics.
The next super-computer might be a protein enriched fluid, with a mosh pit of nano machines at performing the same kinds of dances you observe from an agitated hive of ants on the move. Your brain is like an ant colony doing calculus 3 of multiple variables to render the physics engine that starts and stops when you open and close your eyes.
The instruction set of your brain is solidifying as one of: amplification, inhibition, voting, fusing two proteins together to change their function, cleaving a protein in half to change their function, upsampling and down-sampling of receptor-rafts that float on the surface of your neurons like boats on the water that can multiply or down sample. People thought the neuron was just a signal carrier, not so, the neuron is a small city of machine learning algorithms, and I want to be the first person to untangle all that and show that this can all be recreated in software to create computers that become smart enough to improve their own designs.