AI ALIGNMENT FORUM
AF

Wikitags

AI Capabilities

Edited by plex last updated 29th Aug 2021

AI Capabilities are the growing abilities of AIs to act effectively in increasingly complex environments. It is often compared to to AI Alignment, which refers to efforts to ensure that these effective actions taken by AIs are also intended by the creators and beneficial to humanity.

Subscribe
Subscribe
Discussion0
Discussion0
Posts tagged AI Capabilities
50EfficientZero: human ALE sample-efficiency w/MuZero+self-supervised
gwern
4y
26
48[Paper] Stress-testing capability elicitation with password-locked models
Fabien Roger, Ryan Greenblatt
1y
5
30A small update to the Sparse Coding interim research report
Lee Sharkey, Dan Braun, Beren Millidge
2y
5
32Memorizing weak examples can elicit strong behavior out of password-locked models
Fabien Roger, Ryan Greenblatt
1y
3
64EfficientZero: How It Works
1a3orn
4y
2
41We have achieved Noob Gains in AI
Aniruddha Nrusimha
3y
0
33Interpreting Yudkowsky on Deep vs Shallow Knowledge
Adam Shimi
4y
2
45Evaluations project @ ARC is hiring a researcher and a webdev/engineer
Beth Barnes
3y
6
26Capabilities and alignment of LLM cognitive architectures
Seth Herd
2y
0
42The alignment problem in different capability regimes
Buck Shlegeris
4y
12
33PaLM in "Extrapolating GPT-N performance"
Lukas Finnveden
3y
15
30OpenAI Solves (Some) Formal Math Olympiad Problems
Michaël Trazzi
3y
0
29Will we run out of ML data? Evidence from projecting dataset size trends
Pablo Villalobos
3y
0
35Principles of Privacy for Alignment Research
johnswentworth
3y
20
36The longest training run
Jaime Sevilla, Tamay Besiroglu, Owen D, Anson Ho
3y
2
Load More (15/33)
Add Posts