Ex­plor­ing SAE fea­tures in LLMs with defi­ni­tion trees and to­ken lists

mwatkinsOct 4, 2024, 10:15 PM
37 points
5 comments6 min readLW link

AXRP Epi­sode 37 - Jaime Sevilla on Fore­cast­ing AI

DanielFilanOct 4, 2024, 9:00 PM
21 points
3 comments56 min readLW link

[Question] Seek­ing Solu­tions for Ag­gre­gat­ing Clas­sifier Outputs

Saeid GhafouriOct 4, 2024, 5:39 PM
−1 points
0 comments1 min readLW link

Amoeba roles in tech

Sindhu ShivaprasadOct 4, 2024, 5:25 PM
12 points
0 comments4 min readLW link

LASR Labs Spring 2025 ap­pli­ca­tions are open!

Oct 4, 2024, 1:44 PM
37 points
0 comments4 min readLW link

(Maybe) A Bag of Heuris­tics is All There Is & A Bag of Heuris­tics is All You Need

SodiumOct 3, 2024, 7:11 PM
34 points
17 comments17 min readLW link

Does nat­u­ral se­lec­tion fa­vor AIs over hu­mans?

cdkgOct 3, 2024, 6:47 PM
20 points
1 comment1 min readLW link
(link.springer.com)

What Hayek Taught Us About Nature

Ground Truth DataOct 3, 2024, 6:20 PM
−1 points
6 comments2 min readLW link

Bi­as­ing VLM Re­sponse with Vi­sual Stimuli

Jaehyuk LimOct 3, 2024, 6:04 PM
5 points
0 comments8 min readLW link

AI #84: Bet­ter Than a Podcast

ZviOct 3, 2024, 3:00 PM
56 points
7 comments52 min readLW link
(thezvi.wordpress.com)

[Question] If I have some money, whom should I donate it to in or­der to re­duce ex­pected P(doom) the most?

KvmanThinkingOct 3, 2024, 11:31 AM
36 points
36 comments1 min readLW link

Shut­ting down all com­pet­ing AI pro­jects might not buy a lot of time due to In­ter­nal Time Pressure

ThomasCederborgOct 3, 2024, 12:01 AM
12 points
7 comments12 min readLW link

“25 Les­sons from 25 Years of Mar­riage” by hon­orary ra­tio­nal­ist Fer­rett Stein­metz

CronoDASOct 2, 2024, 10:42 PM
24 points
2 comments1 min readLW link
(theferrett.substack.com)

MIT Fu­tureTech are hiring for a Head of Oper­a­tions role

peterslatteryOct 2, 2024, 5:11 PM
8 points
0 comments4 min readLW link

Can AI Quan­tity beat AI Qual­ity?

Gianluca CalcagniOct 2, 2024, 3:21 PM
2 points
0 comments5 min readLW link

[In­tu­itive self-mod­els] 3. The Homunculus

Steven ByrnesOct 2, 2024, 3:20 PM
75 points
36 comments25 min readLW link

AI Safety Univer­sity Or­ga­niz­ing: Early Take­aways from Thir­teen Groups

agucovaOct 2, 2024, 3:14 PM
26 points
0 comments1 min readLW link

Three main ar­gu­ments that AI will save hu­mans and one meta-argument

avturchinOct 2, 2024, 11:39 AM
8 points
8 comments2 min readLW link

Should we ab­stain from vot­ing? (In non­de­ter­minis­tic elec­tions)

B JacobsOct 2, 2024, 10:07 AM
5 points
6 comments4 min readLW link
(bobjacobs.substack.com)

AI Safety at the Fron­tier: Paper High­lights, Septem­ber ’24

gasteigerjoOct 2, 2024, 9:49 AM
13 points
0 comments7 min readLW link
(aisafetyfrontier.substack.com)

Self-Help Corner: Loop Detection

adamShimiOct 2, 2024, 8:33 AM
88 points
6 comments2 min readLW link
(formethods.substack.com)

The mur­der­ous short­cut: a toy model of in­stru­men­tal convergence

Thomas KwaOct 2, 2024, 6:48 AM
37 points
0 comments2 min readLW link

Switch­ing to a Yamaha P-121 Keyboard

jefftkOct 2, 2024, 2:20 AM
11 points
0 comments2 min readLW link
(www.jefftk.com)

Fore­sight Vi­sion Week­end 2024

Allison DuettmannOct 1, 2024, 9:59 PM
8 points
0 comments1 min readLW link

Happy simulations

FateGrinderOct 1, 2024, 9:05 PM
−5 points
0 comments2 min readLW link

Three Sub­tle Ex­am­ples of Data Leakage

abstractapplicOct 1, 2024, 8:45 PM
172 points
16 comments4 min readLW link

AI Safety Newslet­ter #42: New­som Ve­toes SB 1047 Plus, OpenAI’s o1, and AI Gover­nance Summary

Oct 1, 2024, 8:35 PM
8 points
0 comments6 min readLW link
(newsletter.safe.ai)

Retrieval Aug­mented Genesis

João Ribeiro MedeirosOct 1, 2024, 8:18 PM
6 points
0 comments29 min readLW link

Like­li­hood calcu­la­tion with duobels

Martin GerdesOct 1, 2024, 4:21 PM
4 points
0 comments6 min readLW link

Is Text Water­mark­ing a lost cause?

egor.timatkovOct 1, 2024, 4:20 PM
17 points
13 comments10 min readLW link

In­for­ma­tion dark matter

Logan KiellerOct 1, 2024, 3:05 PM
33 points
4 comments28 min readLW link
(logankieller.substack.com)

Con­ven­tional foot­notes con­sid­ered harmful

dkl9Oct 1, 2024, 2:54 PM
25 points
16 comments1 min readLW link
(dkl9.net)

New­som Ve­toes SB 1047

ZviOct 1, 2024, 12:20 PM
84 points
6 comments32 min readLW link
(thezvi.wordpress.com)

Will AI and Hu­man­ity Go to War?

Simon GoldsteinOct 1, 2024, 6:35 AM
9 points
4 comments6 min readLW link

[Question] AMA: In­ter­na­tional School Stu­dent in China

NoviceOct 1, 2024, 6:00 AM
5 points
0 comments1 min readLW link

AGI Farm

Rahul ChandOct 1, 2024, 4:29 AM
1 point
0 comments8 min readLW link

Why com­par­a­tive ad­van­tage does not help horses

SherrinfordSep 30, 2024, 10:27 PM
101 points
15 comments3 min readLW link

In­tel­li­gence ex­plo­sion: a ra­tio­nal as­sess­ment.

p4rziv4lSep 30, 2024, 9:17 PM
1 point
0 comments1 min readLW link
(docs.google.com)

Peak Hu­man Capital

PeterMcCluskeySep 30, 2024, 9:13 PM
69 points
3 comments5 min readLW link
(bayesianinvestor.com)

Sam Alt­man’s Busi­ness Negging

Julian BradshawSep 30, 2024, 9:06 PM
13 points
0 comments1 min readLW link
(www.bloomberg.com)

In-Con­text Learn­ing: An Align­ment Survey

alamertonSep 30, 2024, 6:44 PM
8 points
0 comments20 min readLW link
(docs.google.com)

Not Just For Ther­apy Chat­bots: The Case For Com­pas­sion In AI Mo­ral Align­ment Research

kenneth_diaoSep 30, 2024, 6:37 PM
2 points
0 comments12 min readLW link

Ex­plor­ing De­com­pos­abil­ity of SAE Features

Vikram_NSep 30, 2024, 6:28 PM
1 point
0 comments3 min readLW link

Knowl­edge Base 1: Could it in­crease in­tel­li­gence and make it safer?

iwisSep 30, 2024, 4:00 PM
−4 points
0 comments4 min readLW link

Point of Failure: Semi­con­duc­tor-Grade Quartz

AnnapurnaSep 30, 2024, 3:57 PM
41 points
8 comments2 min readLW link
(jorgevelez.substack.com)

on bac­te­ria, on teeth

bhauthSep 30, 2024, 3:56 PM
62 points
9 comments6 min readLW link
(bhauth.com)

SB 1047 gets vetoed

ryan_bSep 30, 2024, 3:49 PM
25 points
1 comment1 min readLW link
(www.reuters.com)

Of Birds and Bees

RussellThorSep 30, 2024, 10:52 AM
7 points
9 comments2 min readLW link

A new pro­cess for map­ping discussions

Nathan YoungSep 30, 2024, 8:57 AM
28 points
7 comments6 min readLW link
(open.substack.com)

MATS Alumni Im­pact Analysis

Sep 30, 2024, 2:35 AM
61 points
7 comments11 min readLW link