New, Brief Popular-Level Introduction to AI Risks and Superintelligence

LyleN23 Jan 2015 15:43 UTC

33 points

The very popular blog Wait But Why has published the first part of a two-part explanation/summary of AI risks and superintelligence, and it looks like the second part will be focused on Friendly AI. I found it very clear, reasonably thorough and appropriately urgent without signaling paranoia or fringe-ness. It may be a good article to share with interested friends.

Update: Part 2 is now here.

What links here?

[LINK] Wait But Why—The AI Revolution Part 2 by Adam Zerner (4 Feb 2015 16:02 UTC; 27 points)

LyleN23 Jan 2015 15:43 UTC

33 points

3 comments1 min readLW link Archive

AI Risk AI

wobster109 24 Jan 2015 4:55 UTC
20 points
Agree 100%. Wait But Why is very accessible. Previous posts have focused on the Fermi Paradox, procrastination, sentience/consciousness, religion, and immortality. It reads like a very friendly, very accessible Less Wrong.
Adam Zerner 23 Jan 2015 22:08 UTC
10 points
My one data point:

I don’t have much of a formal background in Math, CS or AI. I know a bit from reading LessWrong and Kurzweil though. There weren’t any core ideas in the article that were new to me… but Wait But Why has a way of explaining things that is really helpful, and my understanding has definitely been solidified after reading that. Plus it was enjoyable.
Artaxerxes 24 Jan 2015 8:57 UTC
8 points

There is some debate about how soon AI will reach human-level general intelligence—the median year on a survey of hundreds of scientists was 2040

This is probably a mistake in the article, because the 2040 median is most likely coming from this survey (www.nickbostrom.com/papers/survey.pdf), and specifically coming from when AI experts were asked what year they could be 50% sure of human level AI coming about. For comparison, the 90% interval was 2075, which paints a very different picture.

Edit: the author has fixed the phrasing on that line