AI Already Thinks. We Just Won't Admit It.
Audio Brief
Show transcript
This episode covers how our language and perception of artificial intelligence will evolve as AI is embodied in physical robots.
There are three key takeaways. First, physical embodiment will eliminate our hesitation to attribute thoughts and intentions to machines. Second, we already naturally attribute cognitive states to software chatbots. Third, mental states should be defined as functional information processing rather than human biological consciousness.
When a physical robot acts on goals, describing its behavior using human vocabulary becomes entirely practical. This shift is already happening as users naturally say a chatbot thinks based on its responses. Reluctance to credit AI with beliefs stems from a flawed definition of mind, whereas cognitive states are simply functional data processing.
As physical robots integrate into society, our language will rapidly adapt to treat artificial systems as genuine cognitive agents.
Episode Overview
- Explores how humanity's language and perception regarding AI "thinking" and "believing" will evolve as AI becomes embodied in physical robots.
- Challenges the philosophical hesitation to attribute genuine beliefs, intentions, and thoughts to artificial systems.
- Demonstrates how we already naturally attribute mental states to software-based systems like chatbots in our everyday language.
- Encourages a reevaluation of what it actually means to possess a "mental state" beyond traditional human-centric, internal-theatre models.
Key Concepts
- Embodiment and Attributing Belief: When AI systems are embodied in physical forms (such as robots), our reluctance to use words like "believe," "intend," and "think" will likely vanish. If a physical entity acts on goals and predictions, it becomes completely practical to describe its behavior using the same vocabulary we use for humans.
- Natural Language Attribution: Even without physical bodies, people already attribute mental states to AI. When a chatbot behaves based on a false premise (e.g., assuming a user is a teenage girl), humans naturally say "it thinks I am a teenage girl," demonstrating that we already treat these systems as having cognitive states.
- The Misconception of Mental States: Our hesitation to attribute thoughts to AI often stems from a flawed human definition of a "mental state." We often imagine an "inner theater" of consciousness, whereas a mental state can be more practically understood as a functional state of processing, predicting, and representing information.
Quotes
- At 0:18 - "It's perfectly reasonable to talk about what the battle robot believes... in the same way as you talk about what a person believes." - This highlights the shifting paradigm where physical, goal-oriented actions by AI make behavioral attribution of beliefs necessary and logical.
- At 0:38 - "Our reluctance to use words like believe and intend and think is gonna disappear once these things are embodied." - Explaining how physical presence and real-world interaction will force us to abandon semantic reservations about AI cognition.
- At 2:05 - "We can attribute mental states to them, but have a completely incorrect model of what it is to have a mental state." - Clarifying that a "mental state" does not require a mystical inner human consciousness, but rather functional information processing.
Takeaways
- Adopt functional language when evaluating AI behavior; recognize that attributing "thinking" or "believing" to a system is often the most accurate and practical way to model its actions.
- Avoid the pitfall of assuming an AI cannot possess "mental states" simply because it lacks human-like biological consciousness or an internal subjective experience.
- Prepare for the societal shift in human-robot interaction by understanding that physical embodiment will dramatically accelerate the human tendency to anthropomorphize and trust AI agency.
