LESSWRONG
LW

1336
Wikitags

Scaling Laws

Edited by riley, plex last updated 18th Jun 2023

Scaling Laws refer to the observed trend that the scaling behaviors of deep neural networks (i.e. how the evaluation metric of interest varies as one varies the amount of compute used for training (or inference), number of model parameters, training dataset size, model input size, or number of training steps) follows variants of power laws. 

External links

  • "Broken Neural Scaling Laws" paper
Scaling laws graph from Scaling Laws for Neural Language Models
Subscribe
Discussion
1
Subscribe
Discussion
1
Posts tagged Scaling Laws
135"Can AI Scaling Continue Through 2030?", Epoch AI (yes)
gwern
1y
4
424chinchilla's wild implications
Ω
nostalgebraist
3y
Ω
128
185What will GPT-2030 look like?
Ω
jsteinhardt
2y
Ω
43
21/r/MLScaling: new subreddit for NN scaling research/discussion
gwern
5y
0
82Thoughts on the Alignment Implications of Scaling Language Models
Ω
leogao
4y
Ω
11
35My ML Scaling bibliography
gwern
4y
9
32Google's new text-to-image model - Parti, a demonstration of scaling benefits
Kayden
3y
4
172o1: A Technical Primer
Ω
Jesse Hoogland
9mo
Ω
19
109Paper: On measuring situational awareness in LLMs
Ω
Owain_Evans, Daniel Kokotajlo, Mikita Balesni, Tomek Korbak, Asa Cooper Stickland, Meg, Maximilian Kaufmann
2y
Ω
17
67Superhuman Coders in AI 2027 - Not So Fast
dschwarz, FutureSearch
4mo
0
63Ethan Caballero on Private Scaling Progress
Michaël Trazzi
3y
2
58Inverse Scaling Prize: Second Round Winners
Ω
Ian McKenzie, Sam Bowman, Ethan Perez
3y
Ω
17
51NVIDIA and Microsoft releases 530B parameter transformer model, Megatron-Turing NLG
Ω
Ozyrus
4y
Ω
36
51[Link] Training Compute-Optimal Large Language Models
Ω
nostalgebraist
3y
Ω
23
50A closer look at chess scalings (into the past)
hippke
4y
14
Load More (15/74)
Add Posts