Your audiobook is waiting…

Human Compatible

Artificial Intelligence and the Problem of Control
Narrated by: Raphael Corkhill
Length: 11 hrs and 38 mins
4.5 out of 5 stars (88 ratings)

$14.95/month after 30 days. Cancel anytime.

OR
In Cart

Publisher's Summary

"The most important book on AI this year." (The Guardian)

"Mr. Russell's exciting book goes deep, while sparkling with dry witticisms." (The Wall Street Journal)

"The most important book I have read in quite some time" (Daniel Kahneman)

"A must-read" (Max Tegmark)

"The book we've all been waiting for" (Sam Harris)

A leading artificial intelligence researcher lays out a new approach to AI that will enable us to coexist successfully with increasingly intelligent machines.

In the popular imagination, superhuman artificial intelligence is an approaching tidal wave that threatens not just jobs and human relationships, but civilization itself. Conflict between humans and machines is seen as inevitable and its outcome all too predictable.

In this groundbreaking audiobook, distinguished AI researcher Stuart Russell argues that this scenario can be avoided, but only if we rethink AI from the ground up. Russell begins by exploring the idea of intelligence in humans and in machines. He describes the near-term benefits we can expect, from intelligent personal assistants to vastly accelerated scientific research, and outlines the AI breakthroughs that still have to happen before we reach superhuman AI. He also spells out the ways humans are already finding to misuse AI, from lethal autonomous weapons to viral sabotage.

If the predicted breakthroughs occur and superhuman AI emerges, we will have created entities far more powerful than ourselves. How can we ensure they never, ever, have power over us? Russell suggests that we can rebuild AI on a new foundation, according to which machines are designed to be inherently uncertain about the human preferences they are required to satisfy. Such machines would be humble, altruistic, and committed to pursue our objectives, not theirs. This new foundation would allow us to create machines that are provably deferential and provably beneficial.

©2019 Stuart Russell (P)2019 Penguin Audio

Critic Reviews

"This is the most important book I have read in quite some time. It lucidly explains how the coming age of artificial super-intelligence threatens human control. Crucially, it also introduces a novel solution and a reason for hope." (Daniel Kahneman, winner of the Nobel Prize and author of Thinking, Fast and Slow)

"A must-read: this intellectual tour-de-force by one of AI's true pioneers not only explains the risks of ever more powerful artificial intelligence in a captivating and persuasive way, but also proposes a concrete and promising solution." (Max Tegmark, author of Life 3.0)

"A thought-provoking and highly readable account of the past, present and future of AI.... Russell is grounded in the realities of the technology, including its many limitations, and isn’t one to jump at the overheated language of sci-fi.... If you are looking for a serious overview to the subject that doesn’t talk down to its non-technical readers, this is a good place to start.... [Russell] deploys a bracing intellectual rigour.... But a laconic style and dry humour keep his book accessible to the lay reader." (Financial Times)

What members say

Average Customer Ratings

Overall

  • 4.5 out of 5 stars
  • 5 Stars
    68
  • 4 Stars
    12
  • 3 Stars
    5
  • 2 Stars
    1
  • 1 Stars
    2

Performance

  • 4.5 out of 5 stars
  • 5 Stars
    65
  • 4 Stars
    9
  • 3 Stars
    4
  • 2 Stars
    0
  • 1 Stars
    1

Story

  • 4.5 out of 5 stars
  • 5 Stars
    56
  • 4 Stars
    12
  • 3 Stars
    6
  • 2 Stars
    1
  • 1 Stars
    3

Reviews - Please select the tabs below to change the source of reviews.

Sort by:
Filter by:
  • Overall
    3 out of 5 stars
  • Performance
    4 out of 5 stars
  • Story
    3 out of 5 stars

Disturbing

Author presents several catastrophic outcomes and then defensively suggests nobody has the right to stop AI research. Also, if the control problem could be solved, we still would not have any reason to believe the increased productivity would not further exacerbate income inequality.

  • Overall
    2 out of 5 stars
  • Performance
    3 out of 5 stars
  • Story
    1 out of 5 stars

Snobby tone of voice for specious arguments

The author makes many slippery slope fallacies and doomsday predictions. And uses the same
faulty reasoning to claim to debunk anti-AI-fears.

There could be a good book in this genre. This is not it.

  • Overall
    5 out of 5 stars
  • Performance
    5 out of 5 stars
  • Story
    5 out of 5 stars

Incisively clear, with breadth to match

In addition to giving a clear-headed overview of the potential, limits, and misuses of contemporary AI technology, Russell does a superb job of articulating the fundamental problem with "the standard model of AI", and how it arises from our intellectual history of trying to engineer "optimal" systems and optimizing machines. The fact that Russell is skeptical of the abilities of current wave of AI and deep learning makes his argument all the more compelling -- the problem with AI isn't rooted in how powerful technology is now, but how we've gone about building it.

Russell proposes instead that we aim to build beneficial machines -- machines that learn human values as they try to assist us -- and is at his best when draws together ideas from philosophy, psychology, economics and computer science to explain both the necessity and difficulty of this aim. A must read for anyone curious about the current state of AI, and concerned about its potential transformative impact on our society.

  • Overall
    5 out of 5 stars
  • Performance
    5 out of 5 stars
  • Story
    5 out of 5 stars

AI, from inception to present day, by world expert

chronicles major problems in AI, prior solutions, and potential future dangers. author is well known, one of the foremost authorities on AI. very listenable historical storytelling.

  • Overall
    1 out of 5 stars
  • Performance
    1 out of 5 stars
  • Story
    1 out of 5 stars

Fluffy with negative marginal value compared to Life 3.0

TLDR: Just read Life 3.0 and spare yourself from this book. If you’ve read Life 3.0 don’t read this book— it will just make you dumber.

I read this book immediately after finishing Life 3.0. I read the whole book.

The author is bombastic and uses statements such as “I’m very confident this will be figured out” rather than hard logic to justify his claims. I wouldn’t mind this so much, except that there are simultaneously very severe holes in his arguments that, even if they don’t completely invalidate them, are sufficient to make his explanations a waste of time to read, especially when considering there are almost zero novel arguments compared to Life 3.0 (I think there is precisely one new concept and it has major holes— see below). Below are some arguments he makes and my problems with them:
1) He argues that giving machines the objective of responding to human preferences and making them uncertain about those preferences will “solve all the problems” associated with misaligned AI.

Hole #1: If the AI is too uncertain about our preferences then it is useless. If the AI is too certain about our preferences then it will have a serious risk of killing us all because it could think that’s what we wanted (or for example wipe out another race of humans— think US-Mexico relations). So clearly we need some way to make sure the AI stays in the “goldilocks” zone with regard to honoring our preferences— but there is absolutely no reason to think that this will be any easier to solve than the original framing of the containment problem. It’s literally just that he reframed the problem of figuring out “how to program the machine so it won’t kill us” to figuring out “what we want the machine to do,” which is basically the same problem.

Hole #2: Related to the above, and part of the reason why “what we want the machine to do” is just as hard to figure out as “how to program the machine so it won’t kill us,” he has a super hand-wavey section about how to “trade off” one human’s preference for another’s, and it gives literally zero reason to think that it is a reconcilable problem. For example, if the British colonists hadn’t wiped out the native Americans, then the life we have today could be technologically less sophisticated. Would the machine deduce that another genocide would be long term beneficial? Would we be ok with that? His response to this is, I’m not even kidding, to say that in some cases “we should put our finger on the scales of justice”— but how is that not *precisely* the loophole problem??

There are several other holes that are, honestly, a bit infuriating. But I think the nail in the coffin for this book is simply that Life 3.0 covers every single concept this guy covers with much, much more rigor and thoughtfulness (with perhaps one exception for the basically completely useless line of reasoning mentioned above). The fact that this guy comes off as a grandpa making blustery, nonrigorous arguments that have major holes doesn’t help.

0 of 4 people found this review helpful