Open Problems in Mechanistic Interpretability: A Whirlwind Tour
A Google TechTalk, presented by Neel Nanda, 2023/06/20
Google Algorithms Seminar - ABSTRACT: Mechanistic Interpretability is the study of reverse engineering the learned algorithms in a trained neural network, in the hopes of applying this understanding to make powerful systems safer and more steerable. In this talk Neel will give an overview of the field, summarise some key works, and outline what he sees as the most promising areas of future work and open problems. This will touch on techniques in casual abstraction and meditation analysis, understanding superposition and distributed representations, model editing, and studying individual circuits and neurons.
About the Speaker: Neel works on the mechanistic interpretability team at Google DeepMind. He previously worked with Chris Olah at Anthropic on the transformer circuits agenda, and has done independent work on reverse-engineering modular addition and using this to understand grokking.
1 view
0
0
17 minutes ago 00:20:29 1
GNOME has money issues, Nvidia driver progress, Asahi can game: Linux & Open Source News
3 hours ago 00:09:25 1
Boss RC-505 Loop Station - Multi-Instrumental Live Looping - Reinhardt Buhr
6 hours ago 00:03:23 1
Nathan James & Arankai - “Paranormal“ (Official Music Video)
7 hours ago 00:28:24 1
Western Media Urges Public: Believe Our Lies, Not Your Own Eyes Regarding China
10 hours ago 00:17:49 1
Viktor Orban: Everyone will agree with me and not with your Left wing lies
10 hours ago 00:03:50 1
Vana - BEG! (Official Music Video)
11 hours ago 00:00:53 1
ПОЛНЫЙ ОБХОД БЛОКИРОВКИ ДИСКОРДА И ЮТУБА ОДНИМ ПРИЛОЖЕНИЕМ БЕЗ ПОТЕРИ КАЧЕСТВА И ПИНГА
18 hours ago 00:02:17 1
ОБХОД БЛОКИРОВКИ DISCORD С ВОЙСОМ | КАК ЗАПУСТИТЬ ДС
21 hour ago 01:06:39 1
1 Hour Live Looping: Building an Entire Ambient Deep House Album from scratch!
22 hours ago 00:03:27 1
Игорь Тальков - I Love You Cover (Lyric Video)
23 hours ago 00:03:33 1
ОБХОД БЛОКИРОВКИ DISCORD И YOUTUBE В РОССИИ 2024! ПОЛНОСТЬЮ РАБОТАЮЩИЙ МЕТОД С ГОЛОСОВЫМ ЧАТОМ
1 day ago 00:03:14 3
DIGGER - THE BOSS (OFFICIAL VIDEO)
1 day ago 00:00:49 4
ОБХОД БЛОКИРОВКИ DISCORD С ГОЛОСОВЫМ ЧАТОМ
1 day ago 00:02:23 1
Siberpunk 2094 - I DON’T LIKE YOU
1 day ago 00:00:00 1
SILENZER - Hurricane (Official Music Video) 2024
1 day ago 00:03:06 1
Set-up HomeLink & open your garage with Cayenne, Macan, Taycan, 911 or Panamera | Tutorial | Spot On
1 day ago 00:03:45 1
Eugy Official - He Called Me (Official Video)
1 day ago 00:08:21 1
Europe’s Problem With The Roma
2 days ago 00:00:40 1
. OST - Sleepless Night + TABS
2 days ago 00:00:00 1
She stopped the Hungarian hurricane Bathory 🥊💥
2 days ago 02:01:04 1
The Rolling Stones live at Soldier Field, Chicago - June 30, 2024 | Full concert + multicam video
2 days ago 00:18:25 1
ИСЦЕЛЯЮЩИЕ КОДЫ/ПРИНЯТИЕ РОДА/БАЛАНС МУЖЧИНЫ И ЖЕНЩИНЫ ВНУТРИ ВАС/ЯЗЫК СВЕТА
2 days ago 00:08:38 1
Friendly Highlights | Sligo Rovers 2-3 Celtic | Celts win in return to Ireland! 🟢⚪