Joar Skalse

Karma: 560

My name is pronounced “YOO-ar SKULL-se”.

I’m a DPhil Scholar at the Future of Humanity Institute in Oxford.

My Criticism of Singular Learning Theory

Joar Skalse19 Nov 2023 15:19 UTC

77 points

56 comments12 min readLW link

Goodhart’s Law in Reinforcement Learning

jacek, Joar Skalse, OliverHayman, charlie_griffin and Xingjian Bai

16 Oct 2023 0:54 UTC

125 points

22 comments7 min readLW link

VC Theory Overview

Joar Skalse2 Jul 2023 22:45 UTC

10 points

2 comments11 min readLW link

How Smart Are Humans?

Joar Skalse2 Jul 2023 15:46 UTC

9 points

19 comments2 min readLW link

Using (Uninterpretable) LLMs to Generate Interpretable AI Code

Joar Skalse2 Jul 2023 1:01 UTC

15 points

9 comments3 min readLW link

Some Arguments Against Strong Scaling

Joar Skalse13 Jan 2023 12:04 UTC

25 points

21 comments16 min readLW link

What kinds of algorithms do multi-human imitators learn?

Chris van Merwijk and Joar Skalse

22 May 2022 14:27 UTC

20 points

0 comments3 min readLW link

Updating Utility Functions

JustinShovelain and Joar Skalse

9 May 2022 9:44 UTC

37 points

6 comments8 min readLW link

Why Neural Networks Generalise, and Why They Are (Kind of) Bayesian

Joar Skalse29 Dec 2020 13:33 UTC

74 points

58 comments1 min readLW link 1 review

[Question] Baseline Likelihood of Long-Term Side Effects From New Drugs?

Joar Skalse27 Dec 2020 13:37 UTC

7 points

2 comments1 min readLW link

Two senses of “optimizer”

Joar Skalse21 Aug 2019 16:02 UTC

35 points

41 comments3 min readLW link

Risks from Learned Optimization: Conclusion and Related Work

evhub, Chris van Merwijk, Vlad Mikulik, Joar Skalse and Scott Garrabrant

7 Jun 2019 19:53 UTC

82 points

5 comments6 min readLW link

Deceptive Alignment

evhub, Chris van Merwijk, Vlad Mikulik, Joar Skalse and Scott Garrabrant

5 Jun 2019 20:16 UTC

117 points

20 comments17 min readLW link

The Inner Alignment Problem

evhub, Chris van Merwijk, Vlad Mikulik, Joar Skalse and Scott Garrabrant

4 Jun 2019 1:20 UTC

103 points

17 comments13 min readLW link

Conditions for Mesa-Optimization

evhub, Chris van Merwijk, Vlad Mikulik, Joar Skalse and Scott Garrabrant

1 Jun 2019 20:52 UTC

83 points

48 comments12 min readLW link

Risks from Learned Optimization: Introduction

evhub, Chris van Merwijk, Vlad Mikulik, Joar Skalse and Scott Garrabrant

31 May 2019 23:44 UTC

183 points

42 comments12 min readLW link 3 reviews

Two agents can have the same source code and optimise different utility functions

Joar Skalse10 Jul 2018 21:51 UTC

11 points

11 comments1 min readLW link