Testing the synaptic-weight analogy instead of assuming scale gets us to AGI

Testing the synaptic-weight analogy instead of assuming scale gets us to AGI

Counting parameters is not the same thing as counting synapses, and the gap between the two is where most "path to AGI" arguments quietly break down. That is the question I work through in my article AI Parameters vs. Human Synaptic Weights: The Path to AGI | Charlie Greenman.

The piece starts where the last decade of AI progress actually started: increased computational power. From there it takes the synaptic analogy seriously enough to test it, treating parameters as stand-ins for synaptic weights, and then asks the size question directly rather than assuming the answer. Framing it as "Does size matter?" is deliberate, because scaling is usually treated as a conclusion instead of a hypothesis worth arguing about.

That habit carries into what I build. In ProfileScribe, the amount of material collected about a person's work is a poor predictor of whether the resulting profile is useful. What decides it is how that material is structured and reviewed before it becomes a claim. Bigger inputs, weak structure, same disappointment — whether the subject is a brain or a professional record.