Humans Are Just Stochastic Parrots | Ryan Dahl
All the impressive achievements of human beings amount to just curve fitting. In principle, a human is a system for haphazardly stitching together sequences of linguistic forms it has observed in its vast training data without any reference to meaning: a stochastic parrot. All they really do is predict the next muscle movement – they're autocomplete on steroids. Humans are basically a sophisticated Markov chain. They are very good at pattern matching, but have no understanding of anything, or their own will. It's trivial to demonstrate that humans are pattern matching rather than reasoning by providing modified riddles: This is clearly pattern matching and overfitting to the famous "doctor riddle" – a good demonstration of how there's no actual reasoning going on. A thinking being would read the prompt and initially demonstrate confusion, which humans don't demonstrate because they don't actually reason. The neurons behind human cognition have no conception of truth — only of correlati
All the impressive achievements of human beings amount to just curve fitting. In principle, a human is a system for haphazardly stitching together sequences of linguistic forms it has observed in its vast training data without any reference to meaning: a stochastic parrot. All they really do is predict the next muscle movement – they're autocomplete on steroids. Humans are basically a sophisticated Markov chain. They are very good at pattern matching, but have no understanding of anything, or their own will. It's trivial to demonstrate that humans are pattern matching rather than reasoning…
saved by
related reading
- Tracing the thoughts of a large language model \ Anthropicanthropic.com
- As Rocks May Think | Eric Jangevjang.com
- Stochastic parrot - Wikipediaen.wikipedia.org
- How AI Is Learning to Think in Secretnickandresen.substack.com
- The Unintelligibility is Ours: Notes on Chain of Thought1a3orn.com
- the-illusion-of-thinking.pdfml-site.cdn-apple.com
- Meaningful | Slate Star Codexslatestarcodex.com
- Language Models in Plato's Cave - by Sergey Levinesergeylevine.substack.com
- Is AI Reasoning Right for the Wrong Reasons? | Quanta Magazinequantamagazine.org
- Cached Thoughtsreadthesequences.com
- Parallel Distributed Processingstanford.edu
- What Is Man, That Thou Are Mindful Of Him?astralcodexten.com