Rhythms of the Brain

Rhythms of the Brain fucked me up in the productive sense. [embed/link: exact edition] I had already been reading machine learning and reinforcement learning papers, where cognition could be rendered as objectives, state, policy, prediction, memory, representations. Then here was the biological substrate being described not as a static wiring diagram but as coordinated activity unfolding across time: oscillations, synchronization, nested rhythms, phase relationships, populations organizing and reorganizing. The brain stopped looking like a bag of neurons and started looking like a dynamical system whose computation was inseparable from timing.

That mattered because timing was conspicuously underrepresented in a lot of the AI I was reading. Deep learning could approximate astonishing mappings while treating the temporal organization of biological systems as mostly irrelevant implementation detail. Neuroscience suggested the opposite possibility: maybe the dynamics are part of the representation, part of the routing, part of how different scales become temporarily coherent. I do not mean I concluded "brain oscillations are the secret to AGI." I mean a whole new space of computational affordances became visible.

The book sent me outward into neuroscience literature. [add specific papers/authors as recovered] Hippocampal dynamics, cortical organization, neural fields, predictive/sensorimotor systems, whatever I could get my hands on. I was fascinated whenever robotics or deep RL independently rediscovered a behavior that biology had already evolved some architecture for, and even more fascinated when somebody tried to import an actual mechanism rather than only stealing the noun. A lot of "brain-inspired" work is marketing: call an attention module attention, call a memory buffer memory, sprinkle a cortical analogy over a conventional network. I wanted the correspondences that constrained implementation.

The more I read, the more scale became a problem. Neurons are real but they are not the only real level. There are cell types, microcircuits, columns depending on how much you buy that framing, regions, long-range networks, neuromodulatory systems, body/environment loops, hemodynamics, electromagnetic fields if those turn out to matter computationally, molecular state below all of it. You can make a model exquisitely accurate at one scale and still fail to preserve the affordances that appear at another. That problem eventually became central to my whole-brain work: how do you represent heterogeneous fields, anatomy, topologies and processes without pretending there is one privileged resolution where "the brain" lives?

There was also a philosophical pressure growing under the science. If subjective experience is produced by physical systems, then whatever structure experience has must be related somehow to structure in those systems. "Consciousness" is too easy a word to use as a fog machine. What are the actual invariants? What distinguishes one experience from another? What does similarity between experiences mean? How do affect, identity, geometry, memory, attention relate? Eventually I encountered The Shape of Experience, which offered a vocabulary around structured experience that clicked with questions I had already been accumulating. [link] I do NOT consider that a final answer. It is another framework I find generative, and if evidence destroys it then good, destroy it.

This is one of the places where my relationship with scientific ideas diverged sharply from my relationship with religion. I can love a framework and still want to kill it experimentally. In fact the love is almost proportional to how much I want to know whether it survives. There is no virtue in preserving a beautiful theory that reality does not instantiate.

Years later this line becomes SuperCognition, whole-brain modeling, SC-WBD, IBM-1, EEG/MEG mappings, individualized dynamics, and increasingly deranged questions about reconstruction and persistence. But I do not want to write those backward into teenage/early-twenties Jacob as if the destination was obvious. It was not. At this point I was just following the graph: AI makes me curious about cognition, cognition makes me curious about brains, brains make me curious about dynamics, dynamics make me curious about experience, and each answer produces a better question.

[embed: annotated book pages] [links: papers that mattered] [embed: early neuroscience notes] [link: Shape of Experience]

And while my head was getting increasingly abstract, there was a giant physical counterexample to abstraction taking shape in the background: I had been trying to build a humanoid robot for years.