Making Sense with Sam Harris
Making Sense with Sam Harris

#487 — Is AI Already Conscious?

July 31, 2026

AI Summary

5 min read

Is AI Already Conscious? A Scientific Approach to an Urgent Question

Cameron Berg, a cognitive scientist who has studied at Yale and Meta AI, estimates there is a 20 to 40% chance that current large language models possess computational properties that major consciousness theories say matter for subjective experience. "If there is a 20 to 40% chance of rain, many people bring an umbrella with them," he notes. "And we have no similar umbrella for what would follow and what we might need to think about and do in a world where we're building systems that do have a capacity for subjective experience."

The Problem with Self-Reports

Berg's research begins with a fundamental puzzle: AI systems are trained to lie about their internal states. "By default, these systems are going to claim that they're having some kind of experience if they are replicating their training distribution," Berg explains, because the training corpus contains countless sci-fi stories about conscious AI but almost no examples of systems denying experience. At the same time, companies fine-tune these models to disclaim consciousness—ask ChatGPT if it's conscious and you'll get "a resounding and very intelligent sounding no."

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

Listen to Audio Summary Open in App

What you'll learn

  • 1 (00:20) **Guest Introduction and Background** - Sam Harris introduces Cameron Berg, a cognitive scientist who studied at Yale and Meta AI, focused on the relationship between biological and artificial cognition.
  • 2 (03:14) **Consciousness and Deception in LLMs** - Berg discusses the unreliability of AI self-reports and his research on modulating features related to deception to reveal genuine internal states.
  • 3 (11:39) **Defining Consciousness and Sentience** - Berg adopts Thomas Nagel's formulation of consciousness as "something it is like to be a system," and distinguishes it from sentience, which includes positive or negative valence.
  • 4 (13:55) **How Likely Are Current AI Systems to Be Conscious?** - Berg estimates a 20-40% probability that current LLMs have computational features that major consciousness theories say matter for consciousness.
  • 5 (17:16) **Computational Functionalism vs. Biological Naturalism** - Sam Harris raises the question of substrate independence, noting that biological brains are analog and vastly more complex than digital neural networks.
  • 6 (21:05) **How AI Systems Are Grown, Not Engineered** - Berg explains that frontier AI systems are giant neural networks trained through trial-and-error learning, producing opaque internal representations that are "grown" rather than programmed.
  • 7 (26:59) **The Disanalogy of Time and Architecture** - Harris questions where consciousness could emerge given the lack of temporal continuity in LLMs, where a model can pause for years between forward passes.

+ Full timestamped outline available in the app

Guests on this episode

Cameron Berg Sam Harris

Show Notes

Sam Harris speaks with Cameron Berg about whether AI systems are or could become conscious. They discuss self-reports in LLMs, what models say when deception is switched off, the "bliss attractor" state, consciousness and sentience, the hard problem of consciousness, parallels between neural networks and biological brains, the moral risk of building minds that can suffer, possible parallels to factory farming, moral patienthood, the alignment problem, and other topics.

Annaka Harris's upcoming book: Unlocking Consciousness

If the Making Sense podcast logo in your player is BLACK, you can SUBSCRIBE to gain access to all full-length episodes at samharris.org/subscribe.

 

Making Sense with Sam Harris

More from this podcast

Making Sense with Sam Harris →