related

• neuro-symbolic-ai
• symbolic_representation

language learning

• learning_vs_acquisition
  • post-wikipedia-input_hypothesis
• comprehensible_input_vs_immersion

(some) language learning related resources

• vid-sch13-lute
• vid-sch13-tools_aside_from_duolingo
• vid-sch13-obsidian_to_learn_german
• vid-sch13-obsidian_to_learn_german
• vid-808cjk-learn_3_languages_at_once

From gemini 2025-06-08

The Role of Language and Culture
​Predictive coding theorists often argue that humans are unique because we offload our predictions into the environment via language and culture.
- ​Words and written languages are essentially engineered symbols that exist outside the brain.
- ​When we perceive a word, it acts as a highly efficient, low-bandwidth constraint on our internal predictive models. 
- Instead of needing to generate complex, bottom-up sensory models to communicate a concept, a discrete external symbol triggers a massive, synchronized top-down cascade of priors in the listener's brain.

From gemini 2025-06-08

Implications for Computational Neuroscience and AI
​The intersection of these two concepts is actively driving the development of Neuro-symbolic AI, which attempts to merge the robust learning and pattern recognition of neural networks with the rigorous logic and reasoning of symbolic systems.
- System 1 vs. System 2: Predictive coding efficiently maps onto Kahneman's "System 1" (fast, automatic, continuous, intuitive inference). Symbolic representation maps onto "System 2" (slow, deliberate, logical, rule-based reasoning).
- Active Inference: Under frameworks like the Free Energy Principle, symbolic reasoning is just another tool the brain uses to minimize long-term prediction error (surprise). Manipulating symbols allows an agent to run highly abstract, counterfactual simulations of the future without the computational cost of simulating every continuous sensorimotor detail.

created on: Sat May 09 2026