Why this blog
I want to keep a record of the questions that arise while reading papers, my interpretations, and the process of turning ideas into projects. This is first a personal research record, and a way for future collaborators to understand my thinking and work.
Architecture Intelligence
My starting point is to ask how network structure, learning algorithms, and the organization of computation can make intelligence interpretable and designable. Neuronal representations, including manifolds, are one aspect of understanding a system; I also want to ask about the processes and mechanisms that produce them.
This is a working perspective, not a proven theory. Notes will distinguish a paper’s findings, my own judgments, and ideas that remain to be tested.
Centered on AI, in conversation with neuroscience
The reading will center on AI papers, with interests in mechanistic interpretability, representation learning, world models, and NeuroAI. Neuroscience informs the questions, without assuming artificial and biological systems are equivalent.
Where things stand
This notebook is just beginning. There are no paper readings or documented projects yet. The opening essay sets out the perspective; future entries will follow the actual reading and work.