There is more than one road to AGI. Taken apart, at least eleven — each from a different disciplinary instinct.
Mathematical. Fractals and entropy. If intelligence is essentially a local structure the universe builds against entropy increase, then mathematics itself is gestating the seed of intelligence.
Biological. Evolution. The brain was not designed; it was filtered by environment. Replicate the filtering, and intelligence grows on its own.
Brain-structure. Copy the cortex. Don't ask why — ask how it's wired. Replicate the wiring diagram in silicon.
Cognitive (four). Empiricism: induced from data. Rationalism: deduced from axioms. Analogy: old solutions mapped onto new problems. Bayesian: a constantly updating belief distribution. Today's large models bet mostly on the first.
Two honest not-yets. Einsteinian thinking: thought experiments, deep conclusions from minimal data — today's models cannot; they devour data. Emotion: we're not even sure of its place in intelligence — noise, or another kind of computation?
The eleven paths need not compete. Real AGI will likely be a confluence of several — just as the brain itself is a trio of evolution, structure, and learning.
Full framework: psyverse.fun
Top comments (0)