Back
Gwern.net – Essays by Gwern Branwen
webgwern.net·gwern.net
Gwern Branwen's personal website contains influential essays on AI scaling, neural networks, and AI safety, including widely-cited work on the neural network scaling hypothesis and tool AI vs. agent AI, making it a notable reference for AI safety researchers.
Metadata
Importance: 62/100homepage
Summary
Gwern.net is the personal website of Gwern Branwen, featuring long-form essays on AI scaling, psychology, statistics, and related topics. It includes influential pieces on the neural network scaling hypothesis, generative AI, and AI safety risks. The site is widely read in AI research and rationalist communities.
Key Points
- •Contains the influential 'NN Scaling Hypothesis' essay, foundational to understanding AI capability growth.
- •Includes 'Tool AIs Want To Be Agent AIs', a notable piece relevant to AI safety and alignment concerns.
- •Covers existential risks, math errors in risk estimation, and embryo selection, relevant to long-term safety thinking.
- •Features empirical self-experiments and meta-science essays promoting rigorous epistemic standards.
- •Broad interdisciplinary coverage including deep learning, decision theory, behavior genetics, and epistemology.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| Gwern Branwen | Person | 52.0 |
1 FactBase fact citing this source
Cached Content Preview
HTTP 200Fetched Apr 7, 202614 KB
Essays · Gwern.net
Skip to main content
Newest
Popular
Notable
Newest: Blog
Statistics
Meta-Science
Decision Theory
Order Statistics
Crypto/Bitcoin
AI: Safety
Generative AI: Image
Generative AI: Fiction
Generative AI: Poetry
Deep Learning
Computer Science
Haskell
Cognition
Psychology
Behavior Genetics
Economics
Economics: Tech
Domestic Cats
Practical
Design
QS: Sleep
QS
Politics
Epistemology
Philosophy
Literary Criticism
Anime
Fiction: Prose
Fiction: Verse
Docs
Docs: Science
Docs: Crypto
Docs: Anime
Docs: NGE
Wikipedia
Reviews
Reviews: Books
Personal
This is the website of Gwern Branwen . I write about AI, psychology, & statistics. I am best known for my writings about AI scaling, poetry & anime neural networks ; darknet markets and Bitcoin ; blinded self-experiments ; and dual n -back & spaced repetition .
Navigation : For information about my site’s philosophy & method, see the About page ; for the website features & implementation, see the Design page ; for information about myself, my use of other websites, and contact information, see the about-me page ; for new pages, see the Changelog ( = new), short blog posts , or new links ; or subscribe to the newsletter ( archives ).
Features : To use dark-mode ( ) or reader-mode ( ), or disable popups ( ), or search the site ( ), use the floating toggle bar in the lower-right upper-right corner ( ); for more, see the help ( ) page. page rewrites titles to condense them and make them shorter, for more uniform solid blocks of text and to improve the layout of multi-column rendering. All the titles in a section should be similar lengths, and not be long enough to trigger wrapping! We want to put our best and most elegant foot forward to readers arriving at Gwern.net for the first time... -->
Newest
“Elegy in a Craneyard”
A Strict Pindaric Meter
“At a Red Light” writing experiment
How Long Would “Weird Al” Survive In A Freezer?
“Apollonian #2: A Poet Teaches AIs”
“Spoilage”
Interview on risktaking & self-experiment
Inkhaven interview on writing ideas
Gwtar: static efficient single-file HTML
Poetry HTML Typesetting
“Apollonian 1: The Counted & Crowned”
Popular
The NN Scaling Hypothesis
Death Note Anonymity
Silk Road 1 Tutorial
DNM Archives (2013–2015)
Spaced Repetition Learning
Commoditize Your Complement
Banner Ads Considered Harmful
Generating Anime Faces
Melancholy of Subculture Society
Ordinary Life Improvements
Notable
Existential Risks & Math Errors
Danbooru2021: Anime Dataset
DNM Archives (2013–2015)
Correlation ≠ Causality
Embryo Selection For Traits
Longevity Meta-Analyses
Tool AIs Want To Be Agent AIs
Internet Research Tips
In Defense Of Inclusionism
Why So Few Matt Levin
... (truncated, 14 KB total)Resource ID:
kb-b3eec02ad0c77dbc