RSS

Matthew Barnett

Karma: 1,173
Page 1

Two ex­pla­na­tions for vari­a­tion in hu­man abilities

Matthew Barnett
25 Oct 2019 22:06 UTC
74 points
18 comments6 min readLW link

Mis­con­cep­tions about con­tin­u­ous takeoff

Matthew Barnett
8 Oct 2019 21:31 UTC
71 points
36 comments4 min readLW link

A sim­ple en­vi­ron­ment for show­ing mesa misalignment

Matthew Barnett
26 Sep 2019 4:44 UTC
64 points
9 comments2 min readLW link

Might hu­mans not be the most in­tel­li­gent an­i­mals?

Matthew Barnett
23 Dec 2019 21:50 UTC
55 points
40 comments3 min readLW link

Will trans­parency help catch de­cep­tion? Per­haps not

Matthew Barnett
4 Nov 2019 20:52 UTC
46 points
5 comments7 min readLW link

Walk­through: The Trans­former Ar­chi­tec­ture [Part 1/​2]

Matthew Barnett
30 Jul 2019 13:54 UTC
30 points
0 comments6 min readLW link

Ex­plain­ing why false ideas spread is more fun than why true ones do

Matthew Barnett
24 Nov 2019 20:21 UTC
30 points
6 comments1 min readLW link

Why Gra­di­ents Van­ish and Explode

Matthew Barnett
9 Aug 2019 2:54 UTC
27 points
9 comments3 min readLW link

Mal­ign gen­er­al­iza­tion with­out in­ter­nal search

Matthew Barnett
12 Jan 2020 18:03 UTC
26 points
11 comments4 min readLW link

Is the term mesa op­ti­mizer too nar­row?

Matthew Barnett
14 Dec 2019 23:20 UTC
25 points
21 comments1 min readLW link

A Primer on Ma­trix Calcu­lus, Part 1: Ba­sic review

Matthew Barnett
12 Aug 2019 23:44 UTC
23 points
2 comments7 min readLW link

In­ner al­ign­ment re­quires mak­ing as­sump­tions about hu­man values

Matthew Barnett
20 Jan 2020 18:38 UTC
23 points
7 comments4 min readLW link

A Sur­vey of Early Im­pact Measures

Matthew Barnett
6 Aug 2019 1:22 UTC
22 points
0 comments8 min readLW link

A Primer on Ma­trix Calcu­lus, Part 2: Ja­co­bi­ans and other fun

Matthew Barnett
15 Aug 2019 1:13 UTC
22 points
6 comments6 min readLW link

Re­think­ing Batch Normalization

Matthew Barnett
2 Aug 2019 20:21 UTC
21 points
5 comments8 min readLW link

Four Ways An Im­pact Mea­sure Could Help Alignment

Matthew Barnett
8 Aug 2019 0:10 UTC
21 points
1 comment8 min readLW link

One Way to Think About ML Transparency

Matthew Barnett
2 Sep 2019 23:27 UTC
20 points
25 comments5 min readLW link

Un­der­stand­ing Batch Normalization

Matthew Barnett
1 Aug 2019 17:56 UTC
19 points
3 comments6 min readLW link