What should I read next? Any AGI safety related material that you can recommend? I've read the following books related (broadly) to AI:

  • Human Compatible: Artificial Intelligence and the Problem of Control,
    by Stuart Russell
  • AI Superpowers: China, Silicon Valley, and the New World Order, by Kai-Fu Lee
  • Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy, by Cathy O'Neil
  • Superintelligence: Paths, Dangers, Strategies, by Nick Bostrom
  • The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World, by Pedro Domingos
  • The Alignment Problem: Machine Learning and Human Values, by Brian Christian
  • Algorithms to Live By: The Computer Science of Human Decisions, by Brian Christian and Tom Griffiths

I find that much (maybe 50%) of what I've read in the above books simply reviews/re-hashes the same handful of concepts (a brief history of AI, what a neural network is, how big data requires a lot of data, what "garbage in garbage out" means, AlexNet was impressive, how impactful AI is and can be, etc.). Several years ago I did some reading/learning about machine learning[1], and I find that I generally don't learn much from reading about AI.[2]

 

  1. ^

    I spent a few months learning python, read various blog posts, did a tiny tutorial to build a very simple toy project with Scikit-learn, and generally developed a decent lay-persons understanding of machine learning. I have a vague familiarity with multiple regression, K nearest neighbors, dimensionality reduction, but I don't have enough of an understanding to describe them for more than a sentence or two, and I definitely don't have enough of an understanding to describe them in a detailed and technical sense.

  2. ^

    The analogy that I am thinking of is that I have of learned the equivalent of the freshmen 100-level course on AI for non-technical people, and all the books that I am reading are also at the 100-level. Are there any non-technical books at the 200-level, or would I have to do a few years of programming in order to be able to understand the 200-level content?

8

0
0

Reactions

0
0
New Answer
New Comment

4 Answers sorted by

Annoyingly, I'm not going to answer your question, but I'm going to ask you a question: having read all of those books, which would you most recommend to a person who was only going to read one book about AI? 

If your answer is 'depends what they're looking for', imagine I'm the one person: my priorities are:
 -a very clear case for why AI might be dangerous, with all the steps laid-out and strongly argued-for, such that I can easily pick out parts where I'm confused or disagree 
-includes relatable everyday examples, both because that will help me understand, and because I'd like some of these at my fingertips so that I can more easily explain AI risk to non-EAs who aren't familiar with it (or aren't familiar with the sorts of risks that EAs worry about). 

Not annoying at all. I'm always happy to share book recommendations. :)

Of all the books I've read related to AI, I think that The Alignment Problem: Machine Learning and Human Values would be the best. I found it far easier to digest. It was written by a journalist rather than an academic, and that shows. The writing style is much smoother than some of the other books on the above list. While it certainly had less detail than Human Compatible or Superintelligence, I think that the increase in "digestibility" far outweighs the decrease in "rigor."

If you are... (read more)

1
Amber Dawn
2y
Thanks, this is really helpful! 

If you are interested in an international politics angle, a relevant recent release is The New Fire: War, Peace, and Democracy in the Age of AI. It will cover some of the same basics but is more sophisticated on the geopolitical dimensions than the books you've listed, none of which were written by people with international security expertise.

Thanks. I have a bit of an international relations background, so I am quite interested in this. Also, I've never heard of this book before, which makes this all the more valuable as a recommendation. :)

If you enjoyed some of the issues raised in Weapons of Math Destruction (which I really enjoyed, as it's an AI book written by an actual developer but focuses on the social issues), you may enjoy going down the regulation/policy rabbithole. None of these are EA books, but I think that's important and in some ways makes them better due to a wider viewpoint.

- Algorithmic Regulation by Karen Yeung and Martin Lodge
This is a great, user-friendly intro to algorithmic regulation, especially because it also explores the how and more importantly why of regulation efforts. Made up of essays from a variety of experts in different areas.

- Robot Rules by Jacob Turner
This is a really good, highly detailed and yet simple introduction to a lot of the legal and regulatory challenges of AI. Written by a very knowledgeable UK-based lawyer. Quite a broad scope but a good foundation of knowledge in this area.

- Advanced Introduction to Law and Artificial Intelligence
This is a bit more 'lawyery' but is still easy to understand for a general reader. It goes via theme which is useful, eg liability, legal personhood, weaponry. Covers multiple Western and some Eastern jurisdictions with examples of how various countries have approached issues.

I also added these as good examples because they don't fall into the 'America is the World' trap that a lot of books do. They focus on global policy and how it interlinks, without just talking about US policy and assuming it's global.

Lovely! Thank you so much for the recommendations. All three of these are books I've never heard of before. Much appreciated.

A lot of people will probably dismiss this due to it being written by a domestic terrorist, but Ted Kaczynski's book Anti-Tech Revolution: Why and How is worth reading. He goes into detail on why he thinks the technological system will destroy itself, and why he thinks it's impossible for society to be subject to rational control. He goes into detail on the nature of chaotic systems and self-propagating systems, and he heavily criticizes individuals like Ray Kurzweil. Robin Hanson critiqued Kaczynski's collapse theory a few years ago on Overcoming Bias. It's an interesting read if nothing else, and has some interesting arguments.

Comments3
Sorted by Click to highlight new comments since: Today at 11:29 AM

What's your goal?

No specific goal; just exploratory learning. I'm curious, I like learning, and I'm looking for book recommendations. I have a moderate preference for learning things that are useful/applicable in relation to EA, but I am not reading about AI in order to become an AI policy expert of a consultant on AI safety or something like that. I'm okay with readings that line up with that, but it isn't something that I am specifically aiming for.

In your case, I highly recommend checking out this eBook called 'The ABCs of AI.' It's a great resource for beginners, offering a straightforward understanding of AI basics and concepts in simple language. As someone who enjoys learning without a specific endpoint, this book provides a solid foundation without delving into overly complex AI stuff. 

I personally read it a while ago before I started delving into complex AI courses and it was so useful to get an overview of the AI technology as a starter.