#222 – Can we tell if an AI is loyal by reading its mind? DeepMind's Neel Nanda (part 1)

315 snips

Sep 8, 2025

Neel Nanda, a researcher at Google DeepMind and a pioneer in mechanistic interpretability, dives into the enigmatic world of AI decision-making. He shares the alarming reality that fully grasping AI thoughts may be unattainable. Neel advocates for a 'Swiss cheese' model of safety, layering various safeguards rather than relying on a single solution. The complexities of AI reasoning, challenges in monitoring behavior, and the critical need for skepticism in research highlight the ongoing struggle to ensure AI systems remain trustworthy as they evolve.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

ANECDOTE

Shutdown Resistance Was Confusion, Not Rebellion

DeepMind investigated so-called self-preservation demos by reading models' chains of thought and found confusion, not genuine scheming.
Simple prompt changes removed shutdown-resistance behavior, supporting the confusion hypothesis.

ADVICE

Start With The Dumbest Tool

Start investigations with the simplest methods and only escalate complexity if needed.
Use chain-of-thought reading and controlled prompt experiments before applying heavy mechanistic tooling.

INSIGHT

Treat Chains Of Thought As Hypotheses

Chains of thought can be informative for hard tasks but are not guaranteed to faithfully reflect internal causation.
Use them as hypothesis generators and validate with causal interventions.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

We don’t know how AIs think or why they do what they do. Or at least, we don’t know much. That fact is only becoming more troubling as AIs grow more capable and appear on track to wield enormous cultural influence, directly advise on major government decisions, and even operate military equipment autonomously. We simply can’t tell what models, if any, should be trusted with such authority.

Neel Nanda of Google DeepMind is one of the founding figures of the field of machine learning trying to fix this situation — mechanistic interpretability (or “mech interp”). The project has generated enormous hype, exploding from a handful of researchers five years ago to hundreds today — all working to make sense of the jumble of tens of thousands of numbers that frontier AIs use to process information and decide what to say or do.

Full transcript, video, and links to learn more: https://80k.info/nn1

Neel now has a warning for us: the most ambitious vision of mech interp he once dreamed of is probably dead. He doesn’t see a path to deeply and reliably understanding what AIs are thinking. The technical and practical barriers are simply too great to get us there in time, before competitive pressures push us to deploy human-level or superhuman AIs. Indeed, Neel argues no one approach will guarantee alignment, and our only choice is the “Swiss cheese” model of accident prevention, layering multiple safeguards on top of one another.

But while mech interp won’t be a silver bullet for AI safety, it has nevertheless had some major successes and will be one of the best tools in our arsenal.

For instance: by inspecting the neural activations in the middle of an AI’s thoughts, we can pick up many of the concepts the model is thinking about — from the Golden Gate Bridge, to refusing to answer a question, to the option of deceiving the user. While we can’t know all the thoughts a model is having all the time, picking up 90% of the concepts it is using 90% of the time should help us muddle through, so long as mech interp is paired with other techniques to fill in the gaps.

This episode was recorded on July 17 and 21, 2025.

Part 2 of the conversation is now available! https://80k.info/nn2

What did you think? https://forms.gle/xKyUrGyYpYenp8N4A

Chapters:

Cold open (00:00)
Who's Neel Nanda? (01:02)
How would mechanistic interpretability help with AGI (01:59)
What's mech interp? (05:09)
How Neel changed his take on mech interp (09:47)
Top successes in interpretability (15:53)
Probes can cheaply detect harmful intentions in AIs (20:06)
In some ways we understand AIs better than human minds (26:49)
Mech interp won't solve all our AI alignment problems (29:21)
Why mech interp is the 'biology' of neural networks (38:07)
Interpretability can't reliably find deceptive AI – nothing can (40:28)
'Black box' interpretability — reading the chain of thought (49:39)
'Self-preservation' isn't always what it seems (53:06)
For how long can we trust the chain of thought (01:02:09)
We could accidentally destroy chain of thought's usefulness (01:11:39)
Models can tell when they're being tested and act differently (01:16:56)
Top complaints about mech interp (01:23:50)
Why everyone's excited about sparse autoencoders (SAEs) (01:37:52)
Limitations of SAEs (01:47:16)
SAEs performance on real-world tasks (01:54:49)
Best arguments in favour of mech interp (02:08:10)
Lessons from the hype around mech interp (02:12:03)
Where mech interp will shine in coming years (02:17:50)
Why focus on understanding over control (02:21:02)
If AI models are conscious, will mech interp help us figure it out (02:24:09)
Neel's new research philosophy (02:26:19)
Who should join the mech interp field (02:38:31)
Advice for getting started in mech interp (02:46:55)
Keeping up to date with mech interp results (02:54:41)
Who's hiring and where to work? (02:57:43)

Host: Rob Wiblin
Video editing: Simon Monsour, Luke Monsour, Dominic Armstrong, and Milo McGuire
Audio engineering: Ben Cordell, Milo McGuire, Simon Monsour, and Dominic Armstrong
Music: Ben Cordell
Camera operator: Jeremy Chevillotte
Coordination, transcriptions, and web: Katy Moore