Go Back
Choose this branch
Choose this branch
meritocratic
regular
democratic
hot
top
alive
246 posts
Interpretability (ML & AI)
Instrumental Convergence
Corrigibility
Myopia
Deconfusion
Self Fulfilling/Refuting Prophecies
Orthogonality Thesis
AI Success Models
Treacherous Turn
Gradient Hacking
Mild Optimization
Quantilization
112 posts
Iterated Amplification
Debate (AI safety technique)
Factored Cognition
Humans Consulting HCH
Experiments
Ought
AI-assisted Alignment
Memory and Mnemonics
Air Conditioning
Verification
10
An Open Agency Architecture for Safe Transformative AI
davidad
11h
11
106
How "Discovering Latent Knowledge in Language Models Without Supervision" Fits Into a Broader Alignment Scheme
Collin
5d
18
70
Can we efficiently explain model behaviors?
paulfchristiano
4d
0
172
The Plan - 2022 Update
johnswentworth
19d
33
31
Paper: Transformers learn in-context by gradient descent
LawrenceC
4d
11
2
(Extremely) Naive Gradient Hacking Doesn't Work
ojorgensen
9h
0
57
You can still fetch the coffee today if you're dead tomorrow
davidad
11d
15
36
[ASoT] Natural abstractions and AlphaZero
Ulisse Mini
10d
1
83
The Singular Value Decompositions of Transformer Weight Matrices are Highly Interpretable
beren
22d
27
58
Multi-Component Learning and S-Curves
Adam Jermyn
20d
24
40
Steering Behaviour: Testing for (Non-)Myopia in Language Models
Evan R. Murphy
15d
16
47
Re-Examining LayerNorm
Eric Winsor
19d
8
64
Engineering Monosemanticity in Toy Models
Adam Jermyn
1mo
6
9
Extracting and Evaluating Causal Direction in LLMs' Activations
Fabien Roger
6d
2
46
Take 9: No, RLHF/IDA/debate doesn't solve outer alignment.
Charlie Steiner
8d
14
50
Notes on OpenAI’s alignment plan
Alex Flint
12d
5
11
Research request (alignment strategy): Deep dive on "making AI solve alignment for us"
JanBrauner
19d
3
79
Beliefs and Disagreements about Automating Alignment Research
Ian McKenzie
3mo
4
84
Rant on Problem Factorization for Alignment
johnswentworth
4mo
48
118
Godzilla Strategies
johnswentworth
6mo
65
3
Alignment with argument-networks and assessment-predictions
Tor Økland Barstad
7d
3
80
Air Conditioner Test Results & Discussion
johnswentworth
6mo
38
116
Supervise Process, not Outcomes
stuhlmueller
8mo
8
106
Preregistration: Air Conditioner Test
johnswentworth
8mo
64
28
A Library and Tutorial for Factored Cognition with Language Models
stuhlmueller
2mo
0
27
Briefly thinking through some analogs of debate
Eli Tyre
3mo
3
18
Ought will host a factored cognition “Lab Meeting”
jungofthewon
3mo
1
132
Debate update: Obfuscated arguments problem
Beth Barnes
1y
21