Go Back
Choose this branch
Choose this branch
meritocratic
regular
democratic
hot
top
alive
51 posts
Machine Learning (ML)
DeepMind
OpenAI
Truth, Semantics, & Meaning
Lottery Ticket Hypothesis
Honesty
Anthropic
Map and Territory
Calibration
52 posts
Interpretability (ML & AI)
AI Success Models
Conservatism (AI)
Principal-Agent Problems
Market making (AI safety technique)
Empiricism
318
DeepMind alignment team opinions on AGI ruin arguments
Vika
4mo
34
223
A challenge for AGI organizations, and a challenge for readers
Rob Bensinger
19d
30
199
Common misconceptions about OpenAI
Jacob_Hilton
3mo
138
123
the scaling “inconsistency”: openAI’s new insight
nostalgebraist
2y
14
114
Understanding “Deep Double Descent”
evhub
3y
51
104
Caution when interpreting Deepmind's In-context RL paper
Sam Marks
1mo
6
100
Gradations of Inner Alignment Obstacles
abramdemski
1y
22
91
Paper: Teaching GPT3 to express uncertainty in words
Owain_Evans
6mo
7
86
Paper: Discovering novel algorithms with AlphaTensor [Deepmind]
LawrenceC
2mo
18
84
Safety Implications of LeCun's path to machine intelligence
Ivan Vendrov
5mo
16
81
Survey of NLP Researchers: NLP is contributing to AGI progress; major catastrophe plausible
Sam Bowman
3mo
6
77
A Bird's Eye View of the ML Field [Pragmatic AI Safety #2]
Dan H
7mo
5
75
SGD's Bias
johnswentworth
1y
16
74
Using GPT-N to Solve Interpretability of Neural Networks: A Research Agenda
Logan Riggs
2y
12
254
A Mechanistic Interpretability Analysis of Grokking
Neel Nanda
4mo
39
183
The Plan - 2022 Update
johnswentworth
19d
33
167
Chris Olah’s views on AGI safety
evhub
3y
38
155
A transparency and interpretability tech tree
evhub
6mo
10
146
Opinions on Interpretable Machine Learning and 70 Summaries of Recent Papers
lifelonglearner
1y
16
134
Deep Learning Systems Are Not Less Interpretable Than Logic/Probability/Etc
johnswentworth
6mo
52
114
How "Discovering Latent Knowledge in Language Models Without Supervision" Fits Into a Broader Alignment Scheme
Collin
5d
18
113
Interpretability/Tool-ness/Alignment/Corrigibility are not Composable
johnswentworth
4mo
8
102
Search versus design
Alex Flint
2y
41
88
A Longlist of Theories of Impact for Interpretability
Neel Nanda
9mo
29
85
A positive case for how we might succeed at prosaic AI alignment
evhub
1y
47
73
Polysemanticity and Capacity in Neural Networks
Buck
2mo
9
69
Engineering Monosemanticity in Toy Models
Adam Jermyn
1mo
6
65
Solving the whole AGI control problem, version 0.0001
Steven Byrnes
1y
7