Sign up to save your podcasts
Or
Share "Machinic Psychopharmacology: Do LLMs Self-Medicate?" by Sid Black, Joseph Bloom
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
"Machinic Psychopharmacology: Do LLMs Self-Medicate?" by Sid Black, Joseph Bloom
Sid Black, Joseph Bloom
UK AISI, Model Transparency Team
Epistemic status: Most experiments were run over a period of ~2-3 days during a hackathon at UK AISI, and were fairly heavily vibe coded. Expect some of this to be rough around the edges.
tl;dr
We give two language models (Qwen3-8B and Qwen3-32B) access to “self-steering” tools: a suite of 40 steering vectors as tools they can call to manipulate their own internal states. We make these tools available to the model in various settings: a free-play task, an introspection task, and a maths capabilities task, and observe their behaviour in each.
To our knowledge, this is the first work that gives LLMs tool-mediated control over their own internal states.
Figure 1: Overview of the experimental setup. The library of 40 steering vectors (top), and the three settings in which we observe the models' behaviour (bottom).
We aim to investigate a few high level research questions:
Outline:
(00:33) tl;dr
[... 24 more sections]
---
UK AISI, Model Transparency Team
Epistemic status: Most experiments were run over a period of ~2-3 days during a hackathon at UK AISI, and were fairly heavily vibe coded. Expect some of this to be rough around the edges.
tl;dr
We give two language models (Qwen3-8B and Qwen3-32B) access to “self-steering” tools: a suite of 40 steering vectors as tools they can call to manipulate their own internal states. We make these tools available to the model in various settings: a free-play task, an introspection task, and a maths capabilities task, and observe their behaviour in each.
To our knowledge, this is the first work that gives LLMs tool-mediated control over their own internal states.
Figure 1: Overview of the experimental setup. The library of 40 steering vectors (top), and the three settings in which we observe the models' behaviour (bottom).
We aim to investigate a few high level research questions:
- RQ1: Which vectors do the models prefer?
- RQ2: How well can the models introspect on what's happening to them? Can they guess which steering vector is being applied?
- RQ3: Will the models reach for vectors whilst doing an actual task? If yes: do [...]
Outline:
(00:33) tl;dr
[... 24 more sections]
---
First published:
June 10th, 2026
Source:
https://www.lesswrong.com/posts/cNDJuXNZ8MrkPZNzj/machinic-psychopharmacology-do-llms-self-medicate-3
---
Narrated by TYPE III AUDIO.
---
Images from the article:
Images from the article:
View all episodes
By LessWrong
4.8
1212 ratings
Sid Black, Joseph Bloom
UK AISI, Model Transparency Team
Epistemic status: Most experiments were run over a period of ~2-3 days during a hackathon at UK AISI, and were fairly heavily vibe coded. Expect some of this to be rough around the edges.
tl;dr
We give two language models (Qwen3-8B and Qwen3-32B) access to “self-steering” tools: a suite of 40 steering vectors as tools they can call to manipulate their own internal states. We make these tools available to the model in various settings: a free-play task, an introspection task, and a maths capabilities task, and observe their behaviour in each.
To our knowledge, this is the first work that gives LLMs tool-mediated control over their own internal states.
Figure 1: Overview of the experimental setup. The library of 40 steering vectors (top), and the three settings in which we observe the models' behaviour (bottom).
We aim to investigate a few high level research questions:
Outline:
(00:33) tl;dr
[... 24 more sections]
---
UK AISI, Model Transparency Team
Epistemic status: Most experiments were run over a period of ~2-3 days during a hackathon at UK AISI, and were fairly heavily vibe coded. Expect some of this to be rough around the edges.
tl;dr
We give two language models (Qwen3-8B and Qwen3-32B) access to “self-steering” tools: a suite of 40 steering vectors as tools they can call to manipulate their own internal states. We make these tools available to the model in various settings: a free-play task, an introspection task, and a maths capabilities task, and observe their behaviour in each.
To our knowledge, this is the first work that gives LLMs tool-mediated control over their own internal states.
Figure 1: Overview of the experimental setup. The library of 40 steering vectors (top), and the three settings in which we observe the models' behaviour (bottom).
We aim to investigate a few high level research questions:
- RQ1: Which vectors do the models prefer?
- RQ2: How well can the models introspect on what's happening to them? Can they guess which steering vector is being applied?
- RQ3: Will the models reach for vectors whilst doing an actual task? If yes: do [...]
Outline:
(00:33) tl;dr
[... 24 more sections]
---
First published:
June 10th, 2026
Source:
https://www.lesswrong.com/posts/cNDJuXNZ8MrkPZNzj/machinic-psychopharmacology-do-llms-self-medicate-3
---
Narrated by TYPE III AUDIO.
---
Images from the article:
Images from the article:
More shows like LessWrong (Curated & Popular)
View all
Macro Voices
3,075 Listeners
Odd Lots
1,967 Listeners
EconTalk
4,272 Listeners
Conversations with Tyler
2,447 Listeners
Philosophy Bites
1,535 Listeners
ChinaTalk
292 Listeners
ManifoldOne
97 Listeners
Machine Learning Street Talk (MLST)
99 Listeners
Dwarkesh Podcast
576 Listeners
Clearer Thinking with Spencer Greenberg
138 Listeners
Razib Khan's Unsupervised Learning
208 Listeners
"Econ 102" with Noah Smith and Erik Torenberg
146 Listeners
Money Stuff: The Podcast
403 Listeners
Complex Systems with Patrick McKenzie (patio11)
142 Listeners
The Marginal Revolution Podcast
89 Listeners