November 08, 2024

“Targeted Manipulation and Deception Emerge when Optimizing LLMs for User Feedback” by Marcus Williams, micahcarroll, Adhyyan Narang, Constantin Weisser, Brendan Murphy

Listen Later

23 minutes

Produced as part of MATS 6.0 and 6.1.

Key takeaways:

Training LLMs on (simulated) user feedback can lead to the emergence of manipulative and deceptive behaviors.
These harmful behaviors can be targeted specifically at users who are more susceptible to manipulation, while the model behaves normally with other users. This makes such behaviors challenging to detect.
Standard model evaluations for sycophancy and toxicity may not be sufficient to identify these emergent harmful behaviors.
Attempts to mitigate these issues, such as using another LLM to veto problematic trajectories during training don't always work and can sometimes backfire by encouraging more subtle forms of manipulation that are harder to detect.
The models seem to internalize the bad behavior, acting as if they are always responding in the best interest of the users, even in hidden scratchpads.

You can find the full paper here.

The code for the project can be [...]

---

Outline:

(00:16) Key takeaways:

(01:31) Why would anyone optimize for user feedback?

(02:25) Simulating user feedback:

(04:52) Harmful behaviors reliably emerge when training with exploitable user feedback

(06:34) Even if most users give good feedback, LLMs will learn to target exploitable users

(08:50) Can this be solved by mixing in harmlessness data?

(09:50) Surely we would detect such harmful behaviors with evals?

(11:17) Can we use another LLM to veto harmful training data?

(13:26) Why does training with the veto not eliminate bad behavior?

(15:47) Chain of Thought Reveals RL-Induced Motivated Reasoning

(18:12) Discussion

(18:15) Model personalization and backdoors as two sides of the same coin.

(19:05) Are our simulated users realistic?

(20:54) What do our results mean for annotator and AI feedback gaming more broadly?

(21:56) Would we expect our method to be a good proxy for future user feedback optimization techniques?

(22:27) Acknowledgements

The original text contained 2 footnotes which were omitted from this narration.

The original text contained 8 images which were described by AI.

---

First published:

November 7th, 2024

Source:

https://www.lesswrong.com/posts/hTFhdrS5TKjxn9Cng/targeted-manipulation-and-deception-emerge-when-optimizing

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

...more

View all episodes

View all episodes

Download on the App Store

Download on the App Store

Get it on Google Play

LessWrong (30+ Karma)

By LessWrong

November 08, 2024

“Targeted Manipulation and Deception Emerge when Optimizing LLMs for User Feedback” by Marcus Williams, micahcarroll, Adhyyan Narang, Constantin Weisser, Brendan Murphy

Listen Later

23 minutes

Produced as part of MATS 6.0 and 6.1.

Key takeaways:

Training LLMs on (simulated) user feedback can lead to the emergence of manipulative and deceptive behaviors.
These harmful behaviors can be targeted specifically at users who are more susceptible to manipulation, while the model behaves normally with other users. This makes such behaviors challenging to detect.
Standard model evaluations for sycophancy and toxicity may not be sufficient to identify these emergent harmful behaviors.
Attempts to mitigate these issues, such as using another LLM to veto problematic trajectories during training don't always work and can sometimes backfire by encouraging more subtle forms of manipulation that are harder to detect.
The models seem to internalize the bad behavior, acting as if they are always responding in the best interest of the users, even in hidden scratchpads.

You can find the full paper here.

The code for the project can be [...]

---

Outline:

(00:16) Key takeaways:

(01:31) Why would anyone optimize for user feedback?

(02:25) Simulating user feedback:

(04:52) Harmful behaviors reliably emerge when training with exploitable user feedback

(06:34) Even if most users give good feedback, LLMs will learn to target exploitable users

(08:50) Can this be solved by mixing in harmlessness data?

(09:50) Surely we would detect such harmful behaviors with evals?

(11:17) Can we use another LLM to veto harmful training data?

(13:26) Why does training with the veto not eliminate bad behavior?

(15:47) Chain of Thought Reveals RL-Induced Motivated Reasoning

(18:12) Discussion

(18:15) Model personalization and backdoors as two sides of the same coin.

(19:05) Are our simulated users realistic?

(20:54) What do our results mean for annotator and AI feedback gaming more broadly?

(21:56) Would we expect our method to be a good proxy for future user feedback optimization techniques?

(22:27) Acknowledgements

The original text contained 2 footnotes which were omitted from this narration.

The original text contained 8 images which were described by AI.

---

First published:

November 7th, 2024

Source:

https://www.lesswrong.com/posts/hTFhdrS5TKjxn9Cng/targeted-manipulation-and-deception-emerge-when-optimizing

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

...more

More shows like LessWrong (30+ Karma)

Making Sense with Sam Harris by Sam Harris

Making Sense with Sam Harris

26,359 Listeners

Conversations with Tyler by Mercatus Center at George Mason University

Conversations with Tyler

2,382 Listeners

The Peter Attia Drive by Peter Attia, MD

The Peter Attia Drive

7,947 Listeners

Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas by Sean Carroll | Wondery

Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas

4,135 Listeners

ManifoldOne by Steve Hsu

ManifoldOne

87 Listeners

Your Undivided Attention by Tristan Harris and Aza Raskin, The Center for Humane Technology

Your Undivided Attention

1,449 Listeners

All-In with Chamath, Jason, Sacks & Friedberg by All-In Podcast, LLC

All-In with Chamath, Jason, Sacks & Friedberg

9,041 Listeners

Machine Learning Street Talk (MLST) by Machine Learning Street Talk (MLST)

Machine Learning Street Talk (MLST)

88 Listeners

Dwarkesh Podcast by Dwarkesh Patel

Dwarkesh Podcast

377 Listeners

Hard Fork by The New York Times

Hard Fork

5,420 Listeners

The Ezra Klein Show by New York Times Opinion

The Ezra Klein Show

15,180 Listeners

Moonshots with Peter Diamandis by PHD Ventures

Moonshots with Peter Diamandis

474 Listeners

No Priors: Artificial Intelligence | Technology | Startups by Conviction

No Priors: Artificial Intelligence | Technology | Startups

121 Listeners

Latent Space: The AI Engineer Podcast by swyx + Alessio

Latent Space: The AI Engineer Podcast

77 Listeners

BG2Pod with Brad Gerstner and Bill Gurley by BG2Pod

BG2Pod with Brad Gerstner and Bill Gurley

455 Listeners