Large
Language
Mind

A process-driven research journal tracing how my understanding of intelligence, cognition and machine learning emerges: from intuition, cross-domain connections and reasoning, to formal concepts.

fig. 1, in three beats
  1. i.Two cells fire at the same moment, again and again.
  2. ii.Each time, both reach a little further toward each other.
  3. iii.Once they touch, the first alone fires the second. After Hebb, 1949.

Atlas

Three lenses the journal keeps returning to. Each note is filed under one or more of them; the latest three show here, the rest are one click away.

aBrain two notes

The organ itself: cells, folds, chemistry, evolution.

II.Bias in biology: defaults shaped by selection I.Biological neurons: analog, plastic, structurally dynamic
bLearning two notes

How connections change with use, in tissue and in circuits.

II.Bias as compression of past experience I.Rosenblatt’s Mark I and its potentiometer weights
cMachine two notes

Graphs, weights and the silicon they run on.

II.Underfitting, priors and the mis-weighting problem I.Taalas HC1: model weights etched into the die

Notes

two notes, showing
II.The Concept of Bias: A Baseline Mechanism for Efficient Intelligence 24 Dec 2025
Bias read through several sciences at once, and the shared pattern behind priors, defaults and compression. filed under cognitive bias, epistemic priors, bias as compression
I.Neuromorphic Neurons vs “von Neumann Neurons” 29 Nov 2025, revised Aug 2026
Why do we simulate neurons in software instead of building them in hardware? From McCulloch and Pitts to Rosenblatt’s Mark I, and on to model-specific silicon. filed under neuromorphic computing, perceptron history, Taalas HC1
iii. not written yet, a new cell will grow here