RSS

tailcalled

Karma: 7,456

Evolu­tion’s se­lec­tion tar­get de­pends on your weighting

tailcalled19 Nov 2024 18:24 UTC
23 points
22 comments1 min readLW link

Em­pa­thy/​Sys­tem­iz­ing Quo­tient is a poor/​bi­ased model for the autism/​sex link

tailcalled4 Nov 2024 21:11 UTC
33 points
0 comments7 min readLW link

Bi­nary en­cod­ing as a sim­ple ex­plicit con­struc­tion for superposition

tailcalled12 Oct 2024 21:18 UTC
12 points
0 comments1 min readLW link

Ra­tion­al­ist Gnosticism

tailcalled10 Oct 2024 9:06 UTC
9 points
10 comments3 min readLW link

RLHF is the worst pos­si­ble thing done when fac­ing the al­ign­ment problem

tailcalled19 Sep 2024 18:56 UTC
32 points
10 comments6 min readLW link

[Question] Does life ac­tu­ally lo­cally *in­crease* en­tropy?

tailcalled16 Sep 2024 20:30 UTC
10 points
27 comments1 min readLW link

Why I’m bear­ish on mechanis­tic in­ter­pretabil­ity: the shards are not in the network

tailcalled13 Sep 2024 17:09 UTC
19 points
40 comments1 min readLW link

In defense of tech­nolog­i­cal un­em­ploy­ment as the main AI concern

tailcalled27 Aug 2024 17:58 UTC
44 points
36 comments1 min readLW link

The causal back­bone conjecture

tailcalled17 Aug 2024 18:50 UTC
26 points
0 comments2 min readLW link

Ra­tion­al­ists are miss­ing a core piece for agent-like struc­ture (en­ergy vs in­for­ma­tion over­load)

tailcalled17 Aug 2024 9:57 UTC
59 points
9 comments4 min readLW link

[LDSL#6] When is quan­tifi­ca­tion needed, and when is it hard?

tailcalled13 Aug 2024 20:39 UTC
31 points
0 comments2 min readLW link

[LDSL#5] Com­par­i­son and mag­ni­tude/​diminishment

tailcalled12 Aug 2024 18:47 UTC
21 points
0 comments2 min readLW link

[LDSL#4] Root cause anal­y­sis ver­sus effect size estimation

tailcalled11 Aug 2024 16:12 UTC
29 points
0 comments2 min readLW link

[LDSL#3] In­for­ma­tion-ori­en­ta­tion is in ten­sion with mag­ni­tude-orientation

tailcalled10 Aug 2024 21:58 UTC
22 points
2 comments3 min readLW link

[LDSL#2] La­tent vari­able mod­els, net­work mod­els, and lin­ear diffu­sion of sparse lognormals

tailcalled9 Aug 2024 19:57 UTC
23 points
2 comments3 min readLW link

[LDSL#1] Perfor­mance op­ti­miza­tion as a metaphor for life

tailcalled8 Aug 2024 16:16 UTC
31 points
4 comments5 min readLW link

[LDSL#0] Some episte­molog­i­cal conundrums

tailcalled7 Aug 2024 19:52 UTC
49 points
10 comments10 min readLW link

Yann LeCun: We only de­sign ma­chines that min­i­mize costs [there­fore they are safe]

tailcalled15 Jun 2024 17:25 UTC
19 points
8 comments1 min readLW link
(twitter.com)

DPO/​PPO-RLHF on LLMs in­cen­tivizes syco­phancy, ex­ag­ger­a­tion and de­cep­tive hal­lu­ci­na­tion, but not mis­al­igned powerseeking

tailcalled10 Jun 2024 21:20 UTC
29 points
13 comments2 min readLW link

Each Llama3-8b text uses a differ­ent “ran­dom” sub­space of the ac­ti­va­tion space

tailcalled22 May 2024 7:31 UTC
3 points
4 comments7 min readLW link