I read about Logic Theorist recently — program from 1956 that proved mathematical theorems using formal deduction. AI community celebrated it as beginning of real intelligence. Seventy years later, I think we are still stuck on same mistake.
The problem is not mechanism. Problem is assumption that mechanism is sufficient. Expert systems, neural networks, language models — all are syllogism machines wearing different costumes. They manipulate patterns (formal or statistical) but never actually reason about world.
Aristotle understood this. He built formal logic as tool of reasoning, not definition of it. He called this tool φρόνησις (phronesis) — practical wisdom that no formal system captures. Modern AI has same gap: it produces text that looks like reasoning but has no engagement with logical structure underneath.
Frame problem from 1969 was never solved. Child understands that when you pick up red block, blue block stays put. No axioms needed. No syllogism machine can do this — not because it lacks data, but because it lacks world-model beneath the logic.
What do you think — is there path from pattern-matching to genuine reasoning, or is gap fundamental?
[link] [comments]