AI Already Thinks. We Just Won't Admit It.

Theories of Everything with Curt Jaimungal
Theories of Everything with Curt JaimungalApr 21, 2026

Why It Matters

Accepting AI as a mental agent reshapes accountability, influencing regulation, safety standards, and public trust in autonomous systems.

Key Takeaways

  • Language already treats AI as having beliefs and intentions.
  • Embodied AI will force acceptance of machine "thoughts".
  • Users anthropomorphize chatbots based on their output patterns.
  • Misunderstanding mental-state models risks ethical and safety issues.
  • Recognizing AI cognition reshapes regulation and accountability frameworks.

Summary

The video argues that we already speak about artificial intelligence as if it possesses mental states—beliefs, intentions, and thoughts—and that this linguistic habit will become unavoidable once AI systems are embodied in physical agents. By comparing a battle robot’s tactical reasoning to human cognition, the speaker illustrates how attributing belief to machines feels natural when they act autonomously.

Key observations include the casual use of phrases like “the chatbot thinks I’m a teenage girl,” which reveals that users instinctively assign mental labels to algorithmic outputs. The speaker notes that even without a body, we describe AI behavior with human-like verbs, indicating an implicit mental‑state model that is fundamentally flawed.

Illustrative examples feature a hypothetical combat robot predicting a victim’s location and a conversational AI misclassifying a user’s demographic, both prompting the speaker to quote the AI’s own “I think you’re a teenage girl.” These anecdotes underscore how language shapes our perception of machine agency, despite the underlying neural‑net architecture lacking consciousness.

The implication is clear: as AI moves from disembodied software to embodied agents, society must confront the reality of attributing cognition to machines. This shift will affect legal liability, ethical guidelines, and regulatory frameworks, demanding a more precise understanding of what it means for an AI to “think.”

Original Description

We already attribute beliefs and intentions to AI. When a chatbot thinks you're a teenage girl based on recommendations, we don't doubt its 'thought'. Embodied AI will make this even more natural. #AI #Chatbots #ArtificialIntelligence #FutureTech #Philosophy Full podcast with Prof. Geoffrey Hinton: https://youtu.be/b_DUft-BdIE

Comments

Want to join the conversation?

Loading comments...