LESSWRONGAI Alignment Writing Day 2018
LW

AI Alignment Writing Day 2018

On 10th July 2018, all attendees of the MIRI Summer Fellows Program were given an entire day to write blogposts to the AI Alignment Forum with ideas they'd been thinking about. These are the 28 posts that resulted, in chronological order.

10Choosing to Choose?

5y

7

13The Intentional Agency Experiment

Alexander Gietelink Oldenziel

5y

5

11Two agents can have the same source code and optimise different utility functions

5y

11

28Conditioning, Counterfactuals, Exploration, and Gears

5y

1

12Probability is fake, frequency is real

Linda Linsefors

5y

7

12Repeated (and improved) Sleeping Beauty problem

Linda Linsefors

5y

5

15Logical Uncertainty and Functional Decision Theory

swordsintoploughshares

5y

4

15A framework for thinking about wireheading

theotherotheralex

5y

4

80Bayesian Probability is for things that are Space-like Separated from You

Scott Garrabrant

5y

22

15A universal score for optimizers

5y

8

15An environment for studying counterfactuals

5y

6

54Mechanistic Transparency for Machine Learning

5y

9

43Bounding Goodhart's Law

5y

2

40A comment on the IDA-AlphaGoZero metaphor; capabilities versus alignment

5y

1

26Dependent Type Theory and Zero-Shot Reasoning

5y

3

22Conceptual problems with utility functions

5y

12

9No, I won't go there, it feels like you're trying to Pascal-mug me

5y

0

12Conditions under which misaligned subagents can (not) arise in classifiers

5y

2

52Complete Class: Consequentialist Foundations

5y

34

20Clarifying Consequentialists in the Solomonoff Prior

5y

16

11On the Role of Counterfactuals in Learning

5y

2

27Agents That Learn From Human Behavior Can't Learn Human Values That Humans Haven't Learned Yet

5y

11

36Decision-theoretic problems and Theories; An (Incomplete) comparative list

5y

0

54 Mathematical Mindset

5y

5

6Monk Treehouse: some problems defining simulation

5y

1

24An Agent is a Worldline in Tegmark V

5y

12

15Generalized Kelly betting

Linda Linsefors

5y

5

16Conceptual problems with utility functions, second attempt at explaining

5y

5