The Interplay of Artificial and Natural Intelligence: A Philosophical Perspective
Clever Hans, Plato, and Large Language Models 🔗
Artificial intelligence (AI) and natural intelligence are interconnected, yet their relationship raises philosophical questions about the nature of intelligence itself. The pursuit of Artificial General Intelligence (AGI) assumes that human cognitive functions can be replicated by machines. However, critics like Hubert Dreyfus argue that human intelligence is far more complex, relying on lived experience and context, which cannot be reduced to algorithms. The paper draws parallels between large language models (LLMs) and the Clever Hans effect, suggesting that both exhibit intelligence not through independent reasoning but through responsive guidance and interaction. The iterative process of prompting in LLMs mirrors Socratic questioning, emphasizing that perceived intelligence stems from external input rather than inherent understanding. This view challenges traditional notions of intelligence as a fixed property, proposing instead that it emerges from relational interactions, underscoring the importance of collaboration in both human and AI contexts.
What is the Clever Hans effect?
The Clever Hans effect refers to the phenomenon where apparent intelligence emerges not from genuine understanding but from sensitivity to external cues. For example, Clever Hans, a horse, seemed to perform arithmetic due to subtle signals from his handler rather than independent reasoning.
How do large language models (LLMs) generate responses?
LLMs generate responses by outputting a probability distribution across their vocabulary, selecting likely words based on patterns learned during training. Their outputs can appear intelligent, but any perceived improvements depend on user guidance rather than true understanding.
What does the paper suggest about the nature of intelligence?
The paper suggests that intelligence should be viewed as an emergent property shaped by interactions and context rather than an inherent trait of individuals or machines. This perspective emphasizes the collaborative nature of intelligence and the importance of guidance in eliciting meaningful responses.