(Doolittle on AI)
Q: Curt: –“How does AI intersect with Metaphor Theory? AI seems very logic based while Metaphor Theory seems to postulate that we humans think in layers and layers of metaphors built up based on our physical experience in the world.”– @patriciamdavis
Metaphor is a category of analogy and analogy is a category of generalization – neurons compete to achieve a coherence of competing networks of generalizations.
AutoAssociation of episodic memory in the hippocampal region seeks analogy for the purpose of identification > prediction > competition > valence > attention (choice).
LLM AI is not logic based (sets), but bayesian probabilism (statistics). Instead, we are having a hell of a problem making it Reason because reason (wayfinding) requires parallel competition and preservation of state. So, Math and Programming ( set logic, procedural logic, finite references and referents) are much much easier than reasoning (episodic logic, infinite referents). And my (our) work requires mathematical decidability, programmatic reasoning, and auto-associative narrating (episodic traversal).
Cheers
CURT
Reply addressees: @patriciamdavis
Source date (UTC): 2025-01-28 18:38:39 UTC
Original post: https://twitter.com/i/web/status/1884310136313069569
Replying to: https://twitter.com/i/web/status/1884306479798440187
Leave a Reply