Book review: Prediction Machines: The Simple Economics of Artificial Intelligence, by Ajay Agrawal, Joshua Gans, and Avi Goldfarb.
Three economists decided to write about AI. They got excited about AI, and that distracted them enough that they only said a modest amount about the standard economics principles that laymen need to better understand. As a result, the book ended up mostly being simple descriptions of topics on which the authors had limited expertise. I noticed fewer amateurish mistakes than I expected from this strategy, and they mostly end up doing a good job of describing AI in ways that are mildly helpful to laymen who only want a very high-level view.
The book’s main goal is to advise business on how to adopt current types of AI (“reading this book is almost surely an excellent predictor of being a manager who will use prediction machines”), with a secondary focus on how jobs will be affected by AI.
The authors correctly conclude that a modest extrapolation of current trends implies at most some short-term increases in unemployment.
Eric Drexler has published a book-length paper on AI risk, describing an approach that he calls Comprehensive AI Services (CAIS).
His primary goal seems to be reframing AI risk discussions to use a rather different paradigm than the one that Nick Bostrom and Eliezer Yudkowsky have been promoting. (There isn’t yet any paradigm that’s widely accepted, so this isn’t a Kuhnian paradigm shift; it’s better characterized as an amorphous field that is struggling to establish its first paradigm). Dueling paradigms seems to be the best that the AI safety field can manage to achieve for now.
I’ll start by mentioning some important claims that Drexler doesn’t dispute:
- an intelligence explosion might happen somewhat suddenly, in the fairly near future;
- it’s hard to reliably align an AI’s values with human values;
- recursive self-improvement, as imagined by Bostrom / Yudkowsky, would pose significant dangers.
Drexler likely disagrees about some of the claims made by Bostrom / Yudkowsky on those points, but he shares enough of their concerns about them that those disagreements don’t explain why Drexler approaches AI safety differently. (Drexler is more cautious than most writers about making any predictions concerning these three claims).
CAIS isn’t a full solution to AI risks. Instead, it’s better thought of as an attempt to reduce the risk of world conquest by the first AGI that reaches some threshold, preserve existing corrigibility somewhat past human-level AI, and postpone need for a permanent solution until we have more intelligence.
Descriptions of AI-relevant ontological crises typically choose examples where it seems moderately obvious how humans would want to resolve the crises. I describe here a scenario where I don’t know how I would want to resolve the crisis.
I will incidentally
ridicule express distate for some philosophical beliefs.
Suppose a powerful AI is programmed to have an ethical system with a version of the person-affecting view. A version which says only persons who exist are morally relevant, and “exist” only refers to the present time. [Note that the most sophisticated advocates of the person-affecting view are willing to treat future people as real, and only object to comparing those people to other possible futures where those people don’t exist.]
Suppose also that it is programmed by someone who thinks in Newtonian models. Then something happens which prevents the programmer from correcting any flaws in the AI. (For simplicity, I’ll say programmer dies, and the AI was programmed to only accept changes to its ethical system from the programmer).
What happens when the AI tries to make ethical decisions about people in distant galaxies (hereinafter “distant people”) using a model of the universe that works like relativity?
Book review: Artificial Intelligence Safety and Security, by Roman V. Yampolskiy.
This is a collection of papers, with highly varying topics, quality, and importance.
Many of the papers focus on risks that are specific to superintelligence, some assuming that a single AI will take over the world, and some assuming that there will be many AIs of roughly equal power. Others focus on problems that are associated with current AI programs.
I’ve tried to arrange my comments on individual papers in roughly descending order of how important the papers look for addressing the largest AI-related risks, while also sometimes putting similar topics in one group. The result feels a little more organized than the book, but I worry that the papers are too dissimilar to be usefully grouped. I’ve ignored some of the less important papers.
The book’s attempt at organizing the papers consists of dividing them into “Concerns of Luminaries” and “Responses of Scholars”. Alas, I see few signs that many of the authors are even aware of what the other authors have written, much less that the later papers are attempts at responding to the earlier papers. It looks like the papers are mainly arranged in order of when they were written. There’s a modest cluster of authors who agree enough with Bostrom to constitute a single scientific paradigm, but half the papers demonstrate about as much of a consensus on what topic they’re discussing as I would expect to get from asking medieval peasants about airplane safety.
Book review: Where Is My Flying Car? A Memoir of Future Past, by J. Storrs Hall (aka Josh).
If you only read the first 3 chapters, you might imagine that this is the history of just one industry (or the mysterious lack of an industry).
But this book attributes the absence of that industry to a broad set of problems that are keeping us poor. He looks at the post-1970 slowdown in innovation that Cowen describes in The Great Stagnation. The two books agree on many symptoms, but describe the causes differently: where Cowen says we ate the low hanging fruit, Josh says it’s due to someone “spraying paraquat on the low-hanging fruit”.
The book is full of mostly good insights. It significantly changed my opinion of the Great Stagnation.
The book jumps back and forth between polemics about the Great Strangulation (with a bit too much outrage porn), and nerdy descriptions of engineering and piloting problems. I found those large shifts in tone to be somewhat disorienting – it’s like the author can’t decide whether he’s an autistic youth who is eagerly describing his latest obsession, or an angry old man complaining about how the world is going to hell (I’ve met the author at Foresight conferences, and got similar but milder impressions there).
Josh’s main explanation for the Great Strangulation is the rise of Green fundamentalism, but he also describes other cultural / political factors that seem related. But before looking at those, I’ll look in some depth at three industries that exemplify the Great Strangulation.
Book review: The Book of Why, by Judea Pearl and Dana MacKenzie.
This book aims to turn the ideas from Pearl’s seminal Causality into something that’s readable by a fairly wide audience.
It is somewhat successful. Most of the book is pretty readable, but parts of it still read like they were written for mathematicians.
History of science
A fair amount of the book covers the era (most of the 20th century) when statisticians and scientists mostly rejected causality as an appropriate subject for science. They mostly observed correlations, and carefully repeated the mantra “correlation does not imply causation”.
Scientists kept wanting to at least hint at causal implications of their research, but statisticians rejected most attempts to make rigorous claims about causes.
Scott at Slate Star Codex made some five year predictions.
I want to encourage a habit of people making long-term forecasts which include quantitative statements, so I’ll give probabilities for most of his predictions, comment on some of his qualitative predictions, then add some predictions of my own.
Book review: Warnings: Finding Cassandras to Stop Catastrophes, by Richard A. Clarke and R.P. Eddy.
This book is moderately addictive softcore version of outrage porn. Only small portions of the book attempt to describe how to recognize valuable warnings and ignore the rest. Large parts of the book seem written mainly to tell us which of the people portrayed in the book we should be outraged at, and which we should praise.
Normally I wouldn’t get around to finishing and reviewing a book containing this little information value, but this one was entertaining enough that I couldn’t stop.
The authors show above-average competence at selecting which warnings to investigate, but don’t convince me that they articulated how they accomplished that.
I’ll start with warnings on which I have the most expertise. I’ll focus a majority of my review on their advice for deciding which warnings matter, even though that may give the false impression that much of the book is about such advice.
[Warning: long post, of uncertain value, with annoyingly uncertain conclusions.]
This post will focus on how hardware (cpu power) will affect AGI timelines. I will undoubtedly overlook some important considerations; this is just a model of some important effects that I understand how to analyze.
I’ll make some effort to approach this as if I were thinking about AGI timelines for the first time, and focusing on strategies that I use in other domains.
I’m something like 60% confident that the most important factor in the speed of AI takeoff will be the availability of computing power.
I’ll focus here on the time to human-level AGI, but I suspect this reasoning implies getting from there to superintelligence at speeds that Bostrom would classify as slow or moderate.
The paper When Will AI Exceed Human Performance? Evidence from AI Experts reports ML researchers expect AI will create a 5% chance of “Extremely bad (e.g. human extinction)” consequences, yet they’re quite divided over whether that implies it’s an important problem to work on.
Slate Star Codex expresses confusion about and/or disapproval of (a slightly different manifestation of) this apparent paradox. It’s a pretty clear sign that something is suboptimal.
Here are some conjectures (not designed to be at all mutually exclusive).