From 2c5859651a736a7b1aa7ffd492bc58cede1e2b4a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Nu=C3=B1o=20Sempere?= Date: Wed, 8 Aug 2018 17:50:58 +0200 Subject: [PATCH] Create ai.md --- maths-prog/ai.md | 71 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 71 insertions(+) create mode 100644 maths-prog/ai.md diff --git a/maths-prog/ai.md b/maths-prog/ai.md new file mode 100644 index 0000000..52c4647 --- /dev/null +++ b/maths-prog/ai.md @@ -0,0 +1,71 @@ +# Artificial Intelligence texts I've read + +In no particular order. Recuerda que probablemente tengas notas más extensas por algún sitio. + +La mayoría extraídos de https://aisafety.com/reading-group/ o https://agentfoundations.org/ + + + +*Cognitive Biases Potentially Affecting Judgement of Global Risks*, Eliezer Yudkowsky + + + +*An AI Race for Strategic Advantage: Rhetoric and Risks*, Sean S. + +*Artificial Intelligence and Its Implications for Future Suffering*, Brian Tomasik + +This guy is a negative utilitarian. + +Is MIRI's work too theoretical? In maths, you can take the supremum of an uncountably infinite set, which you can't do in practice. + +If you have an uncountably infinite number of events, only up to a countably infinite number of them can have nonzero probabilities. + + + +*Impossibility of deducing preferences and rationality from human policy*, Stuart Amstrong + +Too theoretical. In particular, why should I care about maximum regret? + + + +*Refuting Bostrom's Superintelligence Argument*, Sebastian Benthal + +Improving a Bayesian prediction function may have too high a recalcitrance + +|N: I don't really agree. Being able to discriminate between +-0.1db of evidence is probably already a superpower + + + +*There is no fire alarm for AI*, Yudkowsky + + + +*Superintelligence*, Nick Bostrom + + + +*Intelligence Explosion FAQ*, Luke Muehlhauser + + + +*Intelligence Explosion Macroeconomics*. Yudkowsky + + + +*Strategic Implications of Openness in AI development*, Nick Bostrom + + + +*That Alien Message*, Eliezer Yudkowsky + + + +*The Ethics of Artificial Intelligence*, Bostrom and Yudkowsky + + + +*Problem Class Dominance in Predictive Dilemmas*, Daniel Hintze + + + +*Timeless Decision Theory*, Yudkowsky