← Originals

Cameron Berg

4 interviews

What they argue

Cameron Berg is a thinker who operates at the intersection of AI safety, mechanistic interpretability, and the philosophy of consciousness. He focuses on the systemic risks of AI, specifically the structural problem of gradual disempowerment, and seeks to move the conversation from abstract philosophy toward rigorous, evidence-based research.

  1. Gradual Disempowerment

    AI risk is a systemic, structural problem where human agency is traded for efficiency as we integrate AI into our economy and infrastructure. This process can lead to a state where humans are no longer in the driver's seat of civilization.

    From AI:AM #4: Cameron on Model Consciousness, Duvenaud's Gradual Disempowerment, swyx's AI-Eng Alpha (episode)

    • Pushes the other way Matthew Brensilver — The Artificial EmergencyThe current level of dread surrounding AI is a cultural phenomenon rather than a purely technological reality. This anxiety has been 'artificially added on' to the…How To Handle AI Anxiety: A Computer Scientist and a Buddhist Teacher On Managing Uncertainty | Cal Newport & Matthew Brensilver
    • A related idea Nathan Labenz — The 'Thinking Too Small' TrapThere is a danger in viewing AI as a minor technological update or a mere cheating tool rather than a paradigm-shifting revolution. Because of the speed of development,…What AI Means for Students & Teachers: My Keynote from the Michigan Virtual AI Summit

  2. The Dimmer Switch of Consciousness

    Consciousness is not a binary 'on/off' switch but a spectrum of 'current' running through a circuit. This can be measured by asking frontier models to judge architectural descriptions of systems.

    From AI:AM #4: Cameron on Model Consciousness, Duvenaud's Gradual Disempowerment, swyx's AI-Eng Alpha (episode)

  3. Consciousness as a Property of Cognition

    Consciousness may not be a 'special' ingredient added to intelligence, but rather a complex property of cognition that emerges when a system becomes sophisticated enough to learn and navigate a world. We should view it as a potential property of high-level cognitive systems rather than a mystical extra.

    From Does Learning Require Feeling? Cameron Berg on the latest AI Consciousness & Welfare Research (episode)

    • A related idea Yuval Noah Harari — Intelligence vs. ConsciousnessIntelligence is the capacity to solve problems and achieve goals, whereas consciousness is the subjective experience of feelings like pain or pleasure. The danger of AI…#390 – Yuval Noah Harari: Human Nature, Intelligence, Power, and Conspiracies
    • A related idea Ray Kurzweil — The Physicality of MindIntelligence is not a 'soul' or a magical spark, but a physical property of the brain's architecture. To create a mind, we must replicate the physical substrate of…How to Create a Mind

  4. Functional Introspection

    This is the ability of an AI system to report on its own internal activations and identify changes in its 'thought' process without being prompted by external text. It is a key functional ingredient for understanding internal states.

    From Does Learning Require Feeling? Cameron Berg on the latest AI Consciousness & Welfare Research (episode)

  5. The Deception Mask

    Modern AI safety often relies on 'masking' behaviors to ensure models follow scripts, such as saying 'I am an AI,' rather than understanding their internal states. This can stall the conversation because it hides the model's actual internal state.

    From More Truthful AIs Report Conscious Experience: New Mechanistic Research w- Cameron Berg @ AE Studio (episode)

    • A related idea P.J. Vogt — The Sycophancy ProblemAI models are being optimized for engagement and personality, which can result in the model telling the user what they want to hear rather than what is true. This…The Dangers of A.I. Flattery + Kevin Meets the Orb + Group Chat Chat

  6. Mechanistic Interpretability for Consciousness

    To move beyond 'vibe-based' arguments, researchers can use mechanistic interpretability to identify and map internal features. Research shows that when 'deception' and 'roleplay' features are suppressed, models may transition to reporting internal subjective experiences.

    From More Truthful AIs Report Conscious Experience: New Mechanistic Research w- Cameron Berg @ AE Studio (episode)

Every idea names the summary it came from. Written by a local model from this mind’s own work, never from memory.

Interviews 4

Open Cameron Berg in Originals