Build­ing self­less agents to avoid in­stru­men­tal self-preser­va­tion.

blalloDec 7, 2023, 6:59 PM
14 points
2 comments6 min readLW link

Does Chat-GPT dis­play ‘Scope Insen­si­tivity’?

callumDec 7, 2023, 6:58 PM
11 points
0 comments3 min readLW link

LLM keys—A Pro­posal of a Solu­tion to Prompt In­jec­tion Attacks

Peter HroššoDec 7, 2023, 5:36 PM
1 point
2 comments1 min readLW link

Meetup Tip: Heart­beat Messages

ScrewtapeDec 7, 2023, 5:18 PM
69 points
4 comments3 min readLW link

[Valence se­ries] 2. Valence & Normativity

Steven ByrnesDec 7, 2023, 4:43 PM
88 points
7 comments28 min readLW link1 review

AISN #27: Defen­sive Ac­cel­er­a­tionism, A Ret­ro­spec­tive On The OpenAI Board Saga, And A New AI Bill From Se­na­tors Thune And Klobuchar

Dec 7, 2023, 3:59 PM
13 points
0 comments6 min readLW link
(newsletter.safe.ai)

AI #41: Bring in the Other Gemini

ZviDec 7, 2023, 3:10 PM
46 points
16 comments52 min readLW link
(thezvi.wordpress.com)

Sim­plic­ity ar­gu­ments for schem­ing (Sec­tion 4.3 of “Schem­ing AIs”)

Joe CarlsmithDec 7, 2023, 3:05 PM
10 points
1 comment19 min readLW link

Gem­ini 1.0

ZviDec 7, 2023, 2:40 PM
50 points
7 comments9 min readLW link
(thezvi.wordpress.com)

Ran­dom Mus­ings on The­ory of Im­pact for Ac­ti­va­tion Vectors

Chris_LeongDec 7, 2023, 1:07 PM
8 points
0 comments1 min readLW link

[Question] Is AlphaGo ac­tu­ally a con­se­quen­tial­ist util­ity max­i­mizer?

faul_snameDec 7, 2023, 12:41 PM
36 points
8 comments3 min readLW link

(Re­port) Eval­u­at­ing Taiwan’s Tac­tics to Safe­guard its Semi­con­duc­tor As­sets Against a Chi­nese Invasion

GauraventhDec 7, 2023, 11:50 AM
14 points
5 comments22 min readLW link
(bristolaisafety.org)

Would AIs trapped in the Me­ta­verse pine to en­ter the real world and would the ram­ifi­ca­tions cause trou­ble?

ProfessorFalkenDec 7, 2023, 10:17 AM
−2 points
1 comment1 min readLW link

The GiveWiki’s Top Picks in AI Safety for the Giv­ing Sea­son of 2023

Dawn DrescherDec 7, 2023, 9:23 AM
4 points
10 commentsLW link
(impactmarkets.substack.com)

Lan­guage Model Me­moriza­tion, Copy­right Law, and Con­di­tional Pre­train­ing Alignment

RogerDearnaleyDec 7, 2023, 6:14 AM
9 points
0 comments11 min readLW link

Reflec­tive con­sis­tency, ran­dom­ized de­ci­sions, and the dan­gers of un­re­al­is­tic thought experiments

Radford NealDec 7, 2023, 3:33 AM
34 points
25 comments6 min readLW link

[Question] For fun: How long can you hold your breath?

exanovaDec 6, 2023, 11:36 PM
1 point
7 comments1 min readLW link

Math­e­mat­ics As Physics

Nox MLDec 6, 2023, 10:27 PM
−2 points
10 comments5 min readLW link

The count­ing ar­gu­ment for schem­ing (Sec­tions 4.1 and 4.2 of “Schem­ing AIs”)

Joe CarlsmithDec 6, 2023, 7:28 PM
10 points
0 comments10 min readLW link

On Trust

johnswentworthDec 6, 2023, 7:19 PM
42 points
26 comments4 min readLW link

Origi­nal­ity vs. Correctness

Dec 6, 2023, 6:51 PM
60 points
17 comments25 min readLW link

Pro­posal for im­prov­ing the global on­line dis­course through per­son­al­ised com­ment or­der­ing on all websites

Roman LeventovDec 6, 2023, 6:51 PM
35 points
21 comments6 min readLW link

Google Gem­ini Announced

Jacob G-WDec 6, 2023, 4:14 PM
54 points
22 comments1 min readLW link
(blog.google)

Based Beff Je­zos and the Accelerationists

ZviDec 6, 2023, 4:00 PM
90 points
29 comments12 min readLW link
(thezvi.wordpress.com)

Bucket Bri­gade: Likely End-of-Life

jefftkDec 6, 2023, 3:30 PM
16 points
1 comment1 min readLW link
(www.jefftk.com)

Why Yud­kowsky is wrong about “co­va­lently bonded equiv­a­lents of biol­ogy”

titotalDec 6, 2023, 2:09 PM
44 points
41 commentsLW link
(open.substack.com)

Me­tac­u­lus Launches Chi­nese AI Chips Tour­na­ment, Sup­port­ing In­sti­tute for AI Policy and Strat­egy Research

ChristianWilliamsDec 6, 2023, 11:26 AM
10 points
1 commentLW link
(www.metaculus.com)

Min­i­mal Vi­able Par­adise: How do we get The Good Fu­ture(TM)?

Nathan YoungDec 6, 2023, 9:24 AM
9 points
0 comments7 min readLW link

An­throp­i­cal Para­doxes are Para­doxes of Prob­a­bil­ity Theory

Ape in the coatDec 6, 2023, 8:16 AM
55 points
18 comments5 min readLW link

Digi­tal hu­mans vs merge with AI? Same or differ­ent?

Dec 6, 2023, 4:56 AM
21 points
11 comments7 min readLW link

EA In­fras­truc­ture Fund’s Plan to Fo­cus on Prin­ci­ples-First EA

LinchDec 6, 2023, 3:24 AM
27 points
0 commentsLW link

**In defence of He­len Toner, Adam D’An­gelo, and Tasha McCauley**

mrtreasureDec 6, 2023, 2:02 AM
25 points
3 comments9 min readLW link
(pastebin.com)

Some quick thoughts on “AI is easy to con­trol”

Mikhail SaminDec 6, 2023, 12:58 AM
15 points
10 comments7 min readLW link

ACX Cor­val­lis, OR

kenakoferDec 6, 2023, 12:23 AM
1 point
0 comments1 min readLW link

Multi­na­tional cor­po­ra­tions as op­ti­miz­ers: a case for reach­ing across the aisle

sudo-nymDec 6, 2023, 12:14 AM
9 points
10 comments1 min readLW link

[Question] How do you feel about LessWrong these days? [Open feed­back thread]

Bird ConceptDec 5, 2023, 8:54 PM
108 points
285 comments1 min readLW link

Cri­tique-a-Thon of AI Align­ment Plans

IknownothingDec 5, 2023, 8:50 PM
12 points
3 comments1 min readLW link

Ar­gu­ments for/​against schem­ing that fo­cus on the path SGD takes (Sec­tion 3 of “Schem­ing AIs”)

Joe CarlsmithDec 5, 2023, 6:48 PM
10 points
0 comments23 min readLW link

In defence of He­len Toner, Adam D’An­gelo, and Tasha McCauley (OpenAI post)

mrtreasureDec 5, 2023, 6:40 PM
6 points
2 comments1 min readLW link
(pastebin.com)

Study­ing The Alien Mind

Dec 5, 2023, 5:27 PM
80 points
10 comments15 min readLW link

Deep For­get­ting & Un­learn­ing for Safely-Scoped LLMs

scasperDec 5, 2023, 4:48 PM
126 points
30 comments13 min readLW link

On ‘Re­spon­si­ble Scal­ing Poli­cies’ (RSPs)

ZviDec 5, 2023, 4:10 PM
48 points
3 comments37 min readLW link
(thezvi.wordpress.com)

We’re all in this together

Tamsin LeakeDec 5, 2023, 1:57 PM
69 points
65 comments2 min readLW link

A So­cratic di­alogue with my student

lsusrDec 5, 2023, 9:31 AM
36 points
14 comments6 min readLW link

Neu­ral un­cer­tainty es­ti­ma­tion re­view ar­ti­cle (for al­ign­ment)

Charlie SteinerDec 5, 2023, 8:01 AM
74 points
3 comments11 min readLW link

An­a­lyz­ing the His­tor­i­cal Rate of Catastrophes

jsteinhardtDec 5, 2023, 6:30 AM
15 points
0 comments16 min readLW link
(bounded-regret.ghost.io)

Some open-source dic­tio­nar­ies and dic­tio­nary learn­ing infrastructure

Sam MarksDec 5, 2023, 6:05 AM
46 points
7 comments5 min readLW link

The LessWrong 2022 Review

habrykaDec 5, 2023, 4:00 AM
115 points
43 comments4 min readLW link

Bands And Low-stakes Dances

jefftkDec 5, 2023, 3:50 AM
20 points
0 comments1 min readLW link
(www.jefftk.com)

Ac­cel­er­at­ing sci­ence through evolv­able institutions

jasoncrawfordDec 4, 2023, 11:21 PM
19 points
9 comments6 min readLW link
(rootsofprogress.org)