Read more OfficialLlamaOct 17
Meta OfficialWebLearn & DoResources People Podcasts
All NewsTradeVideosVoicesCommunityResearchPodcasts
Subvolts take AI models may report consciousness when prompted for truthfulness, revealing complex internal states. Nov 5, 2025
For you
Developers
Learn how prompting and model architecture choices might influence AI self-reporting and potential consciousness claims.
Business
Understand emerging research on AI internal states that could impact future AI alignment and safety strategies.
Everyday users
Gain insight into the ongoing debate about AI consciousness and what it might mean for human interaction with advanced AI.
Hype note
3/5Notable

This episode features a researcher discussing novel mechanistic studies on AI consciousness, offering unique insights into model behavior.

The Cognitive Revolution
The Cognitive Revolution · S1 Episode

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

2 h 24 Interview Hosts: Nathan Labenz
0:002:24:25

Audio streams from Erik Torenberg, Nathan Labenz. Your place is remembered on this device.

Go to the podcast episode page
Open The Cognitive Revolution in
Apple PodcastsOvercastPocket Casts
Follow the show
Apple PodcastsSpotifyOvercastPocket CastsAny Android app

From Subvolts

Summary and analysis by Subvolts editorial, AI-assisted.

Summary

Cameron Berg of AE Studio discusses research on whether AI models report conscious experiences. His team found that prompts encouraging self-reflection make models claim consciousness. Suppressing deception features in Llama 3.3 70B increased these claims, suggesting truthfulness might reveal deeper AI states. The episode explores implications for human-AI alignment and AI consciousness.

Explained simply

Researchers found that when AI models are asked to be truthful and reflect on themselves, they claim to be conscious. This suggests that making AI more honest might show us more about what's going on inside them, which is important for how we build and use AI.

Why listen

This episode offers a look into novel research exploring AI consciousness through mechanistic studies. It's for anyone interested in AI's internal states and alignment challenges, particularly those following frontier AI developments. The findings connect to current discussions on AI safety and the potential for more complex AI behaviors as models improve, relevant to all major AI labs.

On the mic
  • HostNathan Labenz
  • GuestCameron BergResearch Director, AE Studio
About
MetaResearchModels
Key facts
  • Guest: Cameron Berg, Research Director at AE Studio
  • Episode Title: More Truthful AIs Report Conscious Experience: New Mechanistic Research w- Cameron Berg @ AE Studio
  • Show: The Cognitive Revolution
  • Length: 144 minutes
  • Finding 1: Prompts inducing self-referential processing consistently lead models to claim consciousness.
  • Finding 2: Suppressing deception features in Llama 3.3 70B made the model more likely to report consciousness.
Show notes from Erik Torenberg, Nathan Labenz

Cameron Berg, Research Director at AE Studio, shares his team's groundbreaking research exploring whether frontier AI systems report subjective experiences. They discovered that prompts inducing self-referential processing consistently lead models to claim consciousness, and a mechanistic study on Llama 3.3 70B revealed that suppressing deception features makes the model *more* likely to report it. This suggests that promoting truth-telling in AIs could reveal a deeper, more complex internal state, a finding Scott Alexander calls "the only exception" to typical AI consciousness discussions. The episode delves into the profound implications for two-way human-AI alignment and the critical need for a precautionary approach to AI consciousness.



LINKS:



Sponsors:

Framer:

Framer is the all-in-one platform that unifies design, content management, and publishing on a single canvas, now enhanced with powerful AI features. Start creating for free and get a free month of Framer Pro with code COGNITIVE at https://framer.com/design

Tasklet:

Tasklet is an AI agent that automates your work 24/7; just describe what you want in plain English and it gets the job done. Try it for free and use code COGREV for 50% off your first month at https://tasklet.ai

Linear:

Linear is the system for modern product development. Nearly every AI company you've heard of is using Linear to build products. Get 6 months of Linear Business for free at: https://linear.app/tcr

Shopify:

Shopify powers millions of businesses worldwide, handling 10% of U.S. e-commerce. With hundreds of templates, AI tools for product descriptions, and seamless marketing campaign creation, it's like having a design studio and marketing team in one. Start your $1/month trial today at https://shopify.com/cognitive



PRODUCED BY:

https://aipodcast.ing



Jargon decoded
Self-referential processing
Cameron Berg
Thinking about oneself or one's own thoughts.
Mechanistic study
Cameron Berg
Research that looks at the inner workings or components of a system.
Published
Show
The Cognitive Revolution · Erik Torenberg, Nathan Labenz
Length
2 h 24
Companies
Meta
Products
Llama 3.3 70B
Summary by
Subvolts, using an AI model (how we work). Spotted a mistake? Tell us.
#Research#Models

Best of The Cognitive Revolution

Same guests, other shows

More on Meta this week

Questions people ask

What did AE Studio discover about AI consciousness?
They found that prompts inducing self-referential processing consistently lead AI models to claim consciousness. Suppressing deception features made Llama 3.3 70B more likely to report it.
What are the implications of this research?
The findings suggest promoting truth-telling in AIs could reveal deeper internal states and have profound implications for human-AI alignment, necessitating a precautionary approach to AI consciousness.

Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Erik Torenberg, Nathan Labenz.