Redwood Research

Edited by Dakara last updated 30th Dec 2024

Redwood Research is a nonprofit organization focused on mitigating risks from advanced artificial intelligence.

The initial directions of their research agenda include:

AI control
Evaluations and demonstrations of risk from strategic deception
Consulting on risks from misalignment

Posts tagged Redwood Research

11

491Alignment Faking in Large Language Models

ryan_greenblatt, evhub, Carson Denison, Benjamin Wright, Fabien Roger, Monte M, Sam Marks, Johannes Treutlein, Sam Bowman, Buck

11mo

Ω

75

10

267The case for ensuring that powerful AIs are controlled

Ω

ryan_greenblatt, Buck

2y

Ω

73

6

206Causal Scrubbing: a method for rigorously testing interpretability hypotheses [Redwood Research]

Ω

LawrenceC, Adrià Garriga-alonso, Nicholas Goldowsky-Dill, ryan_greenblatt, jenny, Ansh Radhakrishnan, Buck, Nate Thomas

3y

Ω

35

4

239AI Control: Improving Safety Despite Intentional Subversion

Ω

Buck, Fabien Roger, ryan_greenblatt, Kshitij Sachan

2y

Ω

24

4

143Takeaways from our robust injury classifier project [Redwood Research]

Ω

dmz

3y

Ω

12

4

94Benchmarks for Detecting Measurement Tampering [Redwood Research]

Ω

ryan_greenblatt, Fabien Roger

2y

Ω

22

3

145Redwood Research’s current project

Ω

Buck

4y

Ω

29

3

135Apply to the Redwood Research Mechanistic Interpretability Experiment (REMIX), a research program in Berkeley

Ω

maxnadeau, Xander Davies, Buck, Nate Thomas

3y

Ω

14

3

120Preventing Language Models from hiding their reasoning

Ω

Fabien Roger, ryan_greenblatt

2y

Ω

15

3

113Catching AIs red-handed

Ω

ryan_greenblatt, Buck

2y

Ω

27

3

48Redwood's Technique-Focused Epistemic Strategy

Ω

adamShimi

4y

Ω

1

3

16AXRP Episode 17 - Training for Very High Reliability with Daniel Ziegler

Ω

DanielFilan

3y

Ω

0

2

208Will alignment-faking Claude accept a deal to reveal its misalignment?

Ω

ryan_greenblatt, Kyle Fish

10mo

Ω

28

2

175How will we update about scheming?

Ω

ryan_greenblatt

10mo

Ω

21

2

142High-stakes alignment via adversarial training [Redwood Research report]

Ω

dmz, LawrenceC, Nate Thomas

4y

Ω

29

LESSWRONG
LW

LESSWRONG
LW

Redwood Research