Architectural Principles & Computation in the Brain

in progress

Brain Development

In Systems Neuroscience we often talk about Efficient Coding, but perhaps equally important is the problem of “Efficient Wiring”. A biological information processing system can only become as complex as its developmental pathways and trajectories afford. I believe Development holds part of the key to understanding the components list of the brain, and that ontological and phylogenetic relationships can provide surprising clues into mature function. Furthermore, it focuses on stages of life where learning rate and learning rules are fundamentally different, and when many priors (expectations about the world) are first formed. Any path towards Artificial General Intelligence will in my opinion require a solid understanding of developmental processes.

Extracellular Neurophysiology

Extracellular probes allow neuroscientists to record from virtually any brain region in freely-moving animals at the temporal and spatial scale used in neural computation (though I’m convinced a lot of computation also happens molecularly). For these reasons - and the sheer beauty of the spike - extracellular neurophysiology is my favourite technique in neuroscience. It is also an area of great application and importance, since it constitutes the technology behind most Brain-Computer Interfaces. Understanding and advancing Extracellular Neurophysiology technology and analysis tools will help us learn more about the brain and make progress in curing neurological disease and trauma, as well as pave the road for neural augmentation. This area is gathering momentum once again with startups (for example CTRL-labs and Neurable) and big tech (Facebook, Elon Musk’s Neuralink) becoming interested and publishing fantastic advances.

In my postdoctoral work with Adam Kampff we collected a ground-truth dataset where I recorded the same neuron in vivo using Neuropixels probes and patch-clamp. This dataset and others like it are used as benchmarking tools for improving and developing spike-sorting algorithms, which are the analysis tools we use on raw extracellular data to separate and identify the impulses fired by individual neurons. I am actively conducting research in this area and spoke recently about this at an excellent conference in Edinburgh.

Science Structure and Policy

I believe the measures that would most accelerate the rate of scientific progress have to do with changes in organisational structure, publishing and transparency policies, rather than any technical or technological advancements. Strictly speaking, these are not areas I do research on, but they’re subjects I’m very passionate about and I post thoughts about them regularly on the blog. I have previously talked about these topics in meetings and discussion panels and am generally quite open to discussing them.