Fetching the paper…

Reading the bibliography…

2017

Low Impact Artificial Intelligences

Armstrong, Stuart, Levinstein, Benjamin

Original
PDF
Post on X

Understand

There are many goals for an AI that could become dangerous if the AI becomes superintelligent or otherwise powerful.

Much work on the AI control problem has been focused on constructing AI goals that are safe even for such AIs.
This paper looks at an alternative approach: defining a general concept of `low impact'.
The aim is to ensure that a powerful AI which implements low impact will not modify the world extensively, even if it is given a simple or dangerous goal.

Low Impact Artificial Intelligences · Around

Built on

Fact, fiction, and forecast
Nelson Goodman · 1983
Earlier work this paper cites.
The symbol grounding problem
Stevan Harnad · 1990
Earlier work this paper cites.
Game theory, maximum entropy, minimum discrepancy, and robust bayesian decision theory
P.D. Grünwald and A.P. Dawid · 2004
Earlier work this paper cites.
Strictly proper scoring rules, prediction, and estimation
Tilmann Gneiting and Adrian E. Raftery · 2007
Earlier work this paper cites.
Basic ai drives
S. Omohundro · 2008
Earlier work this paper cites.

Similar

Whole brain emulation. a roadmap
Anders Sandberg and Nick Bostrom · 2008
Cited alongside, same era.
Artificial intelligence as a positive and negative factor in global risk
Eliezer Yudkowsky · 2008
Cited alongside, same era.
Learning what to value
Daniel Dewey · 2011
Cited alongside, same era.
Thinking inside the box: Controlling and using an oracle ai
Stuart Armstrong, Anders Sandberg, and Nick Bostrom · 2012
Cited alongside, same era.

Then

Leakproofing the singularity: artificial intelligence confinement problem
Roman V. Yampolskiy · 2012
Later among the works it cites.
Definability of truth in probabilistic logic
Paul Christiano, Eliezer Yudkowsky, Marcello Herreshoff, and Mihaly Barasz · 2013
Later among the works it cites.
Superintelligence: Paths, dangers, strategies
Nick Bostrom · 2014
Later among the works it cites.
Motivated value selection for artificial agents
Stuart Armstrong · 2015
Later among the works it cites.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…