Books reviewed:
- The Singularity Is Near: When Humans Transcend Biology by Ray Kurzweil (2006)
- The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies by Erik Brynjolfsson and Andrew McAfee (2014)
- The Economic Singularity: Artificial intelligence and the death of capitalism by Calum Chace (2016)
Introduction
One of the biggest challenges facing our society now and in the near future, outside of global warming and fascism, is the automation of the labor force due to nearly science-fiction level advances in artificial intelligence.
People tend to separate out “artificial intelligence” from “automation” but the two go hand in hand. Automation, from simple manufacturing robots to driverless cargo trucks, relies on advanced computer intelligence, which is another way of saying “artificial” intelligence. In the computer science field it’s a lot more complicated and nuanced than that*, but to most of us, this all boils down to “computers that can think for themselves at some level.” A roomba might not strike anyone as smart but it is a machine using “sensory input” to “make decisions” in order to get its job done, and that could pretty much be said of any human who has worked a production line job at a manufacturing plant over the last 200 years. Technologically speaking, manufacturing robots are a step up from roombas, and driverless cars are many steps up from that, but they are all on the “intelligent machine” spectrum even if they are, as yet, vastly inferior to human intelligence.**
For now.
There is no tipping point, but rather a long seismic shift that we are living in the middle of. There currently exist computers/programs that are, by most measures, smarter than a lot of humans. At the same time, teaching a robot how to fold a towel as quickly as a human does is a hurdle not yet cleared.
For now.
But those measures are constantly in motion, and all anyone has to do is watch a video from Boston Dynamics to know that things are changing, and fast. Autonomous cars from companies like Alphabet (Google) have already clocked more than a million miles of travel time on public roads. Capitalism gives all industries a strong motivation for replacing faulty, clumsy humans with robots who never sleep, eat, complain, or unionize. Manufacturers can hire less than a dozen people to oversee robots doing work that used to take hundreds of people to accomplish, and companies like Uber are actively planning to replace all their “contracted” drivers with autonomous cars within a decade. We have robots building robots and computer software coding more software. Whether it takes 20 years or 200, humanity’s time at the top of the evolutionary chain is very limited.
What does it mean to have machines take over human jobs and, possibly, humanity itself? In choosing the books to read on this topic, I wanted to dig down to the foundations of the issue while keeping a broad, global scope in mind. That narrowed the field a bit, but I could not not read the modern classic by Kurzweil, whose computer science cred is the stuff of legend and is best known by the general populace for helping to popularize the concept of a “technological singularity” (the point in time at which a machine/computer/robot becomes self-aware and then swiftly moves to becoming a god-like super-intelligence).
His work coalesces much of what has been written earlier, so instead of marching backwards I take us forward and review recent books focusing on issues that will directly impact humanity as a whole (as opposed to more esoteric, philosophical writings that have also come out of late).
I’ve had a hard time deciding a good reading order for these three books. In the end I think the chronological order of when they were written/published is probably the best choice, as it does give a bit of perspective on the current in-progress historical arc going on. The earliest, Kurzweil’s seminal modern classic The Singularity is Near, is a deep dive into AI and speculation about the future but is tempered ten years on by what has really happened, technologically and politically. So, that’s where we are starting.
Disclosure: I actually read the last book first, after reading Brad Wardell’s alarming and alarmist blog post, Economic Singularity: the Gods and the Useless and deciding it was time to educate myself on the overall topic. Chace’s book seemed to fit the bill as an easily accessible overview, and I figured I could just dismiss any of his more outlandish claims with a spot of googling or two.
I couldn’t, and because I couldn’t, I went back to square one with Kurzweil and followed with Brynjolfsson and McAfee. I think the reading order I suggest here makes more sense.
Book 1
The Singularity Is Near: When Humans Transcend Biology by Ray Kurzweil (2006)
This book is required reading for anyone interested in artificial intelligence and the future of humanity, if only because it lays out the basic speculations and is also referred to in all the other current literature about the subject. If you are investigating the topic of artificial intelligence, you cannot escape Ray Kurzweil.
This is not an easy book because Kurzweil has the mind of a compsci geek and revels in the technological minutiae of his explanations, while on the other extreme dives into the fantastical with a futurism so divorced from humanity as to make most readers uncomfortable. If you are not familiar with the basic premise of “transhumanism” (or are not a reader familiar with science fiction genres like cyberpunk) the speculation Kurzweil puts forth might seem shocking. To me, though, that’s the least interesting part of the book.
Kurzweil spends a lot of time on explaining the pace of technological development in order to set the scene for his premonitions, and that’s where the meat of the book is, in my opinion. He delves into Moore’s Law and exponential growth, both of which are something like buzzword phrases in the tech field, but also veers off into evolution, life cycles of technology, the human brain, and the limits of computation as we know them. It’s a layer cake of science and technology which is quite delicious and worth savoring.
There are whole websites devoted to proving and/or debunking Kurzweil’s predictions, but it’s enough to look at his thoughts about the revolutionary nature of carbon nanotubes in computer evolution to realize that the value of his predictions is in the eventuality of them, not the time span he gives for actualization. Carbon nanotubes are, ten years after publication of this book, still rife with amazing potential…but that’s about it. There have been some major scientific stumbling blocks that appeared along the way to the carbon nanotube revolution that have slowed the pace of progress in that field. Kurzweil was super optimistic about them and super dead wrong about how soon they would be in production, but he is not wrong about how revolutionary they will be when technological obstacles are finally cleared.
This is how the whole books goes, to be honest. Ignore his predictions for time spans, or maybe just double them, but don’t throw out the baby with the bathwater. Kurzweil is not wrong about anything here, he’s just over enthusiastic. That should give you pause, because Kurzweil talks about how humanity must embrace our transhumanist future in conjunction with a superintelligent AI, or become extinct. This might sound shocking to people who have never given it much thought.
I am reminded of the civilization of Frank Herbert’s Dune universe, where AI was outlawed because it was believed that robots had turned against humanity (no matter which version of the history behind the Butlerian Jihad you prefer, it’s basically the same premise), thus complicating technological advancement while degrading cultural advancement — humans did not, in fact, become extinct, but they paid a price for that survival and it serves as a warning bell. Kurzweil’s arguments about the eventuality of the singularity make it clear that this is a choice humanity will have to face at some point soon, and most of us are hardly ready (philosophically, spiritually, or personally) to make it.
This book will make you think long and hard about humanity’s relationship to technology, no matter what your opinion of Kurzweil’s conclusions.
Book 2
The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies by Erik Brynjolfsson and Andrew McAfee (2014)
This book is much more accessible than Kurzweil’s to the general reader but covers some of the same ground as The Singularity is Near, except for the more outrageous transhumanism predictions and superintelligent AI theories. Their focus is not on the esoteric science of it; they are economists, and the economy is ultimately what concerns them.
The writers quickly cover material that Kurzweil spilled a lot of electric ink over, such as exponential growth and Moore’s Law, and jump right into parsing out short term advances and their impact. They reference a lot of academic studies, putting some force behind their own conjectures, and spend time describing visits they took to different labs and companies in order to look at what is actually being done “on the ground” so-to-speak. This is where the book is the most interesting, because the combination of researchers who know the data and the trends facing off with developers and entrepreneurs presents a very holistic view of what is actually going on.
The result is a much more measured and hesitant presentation of, basically, everything Kurzweil already said nearly ten years earlier. The difference is that they were the first to posit an “economic singularity,” which loosely is defined as the point at which the majority of human jobs will be displaced by automation/computers/robots. This is generally believed to be a point that will be reached some time in advance of the technological singularity that Kurzweil is focused on, and thus is a more urgent concern even it if is not quite as important.
Of particular interest is their overview of the advancements to autonomous cars. Kurzweil did not spend much time on autonomous cars, aside from predicting they would advance quickly. Given that he was writing in the same year as the infamous DARPA Challenge of 2005 that failed to see even a single autonomous car cross the finish line, that was viewed as one of his more outlandish proclamations. As Brynjolfsson and McAfee point out, he was actually too cautious on that one, and currently it seems like at least partially autonomous cars will be sold by most major car manufacturers by 2020. Brynjolfsson and McAffee look closely at how, exactly, that happened over the course of one decade when most experts were claiming it would take 20+ years to even develop workable prototypes. This is a great example of why this book is worth reading: it breaks down the history of expectations and matches that against current progress, then pushes forward a little into what’s on the horizon for that industry. They do this across a number of industries.
This gives the book a very balanced, reasonable perspective that is the opposite of shocking or alarmist. Which, unfortunately, is its greatest weakness. In trying to assuage all fears and present the impact of these, frankly breathtaking, technological changes as nothing more than a slight variation to the status quo, the authors spend the last part of the book riding the middle ground of compromise and underselling the dangers. They don’t want to scare the conservatives or progressives, so take great pains to show how all of this will not change or disrupt capitalism in any possible way while also projecting that the end of labor as we know it will result in the need for some kind of government intervention. They hem and haw over universal basic income (UBI) concepts (as there is more than one way to implement it), finally falling in with the negative tax proposal as the least disruptive. This is a topic where the answers will never please everyone, but the authors try their best.
It’s nice to believe that our current society can move into this glorious post-singularity future with very little change, but I believe it is dead wrong. My take-away from Kurzweil’s book was that things are changing in fundamental ways, and Brynjolfsson and McAfee don’t present any real evidence contrary to that. Most of what they discuss reinforces it, in fact, despite their attempts to hold back the throttle.
So for book three, we push it to the max.
Book 3
The Economic Singularity: Artificial intelligence and the death of capitalism by Calum Chace (2016)
Chace has a series of self-published books on the topics if AI, robotics, and automation, and seems to mostly be a self-identified expert in the issue. Usually that’s a sign to steer clear, so I approached this book with a grain of salt firmly held. As I stated in the introduction, I read this book right after reading Wardell’s blog post, and it seemed pretty outrageous. I came back to it after finishing the other books, and I realized with come chagrin that I still couldn’t dismiss anything in Chace’s book.
He’s done his homework, and it shows.
He claims to have 30+ years as a journalist and businessman under his belt, and while I can’t prove or disprove that, I’m inclined to believe it. His analyses have the hard-hitting impact that comes from such a background, as opposed to the more “hedge your bets” style that academics go for (such as Brynjolfsson and McAfee, as I noted above). His book reads more like a well researched business plan than a thoughtful treaty on philosophical quandaries.
If you’ve read the other two books first (as I didn’t!) then not much here comes as a surprise. However instead of spending time explaining exponential growth and Moore’s Law in detail first, he folds that into an historical overview of technological change, devoting initial chapters to the industrial, information, and automation revolutions in turn. This is pretty crucial to his argument about the economic singularity, and he spends the majority of the book explaining why this is important. And, indeed, it is very important.
As the title makes clear, this book is not overly interested in the technological singularity and what that may mean for the future of humanity. Chace is singularly focused (heh) on the economic singularity, and what that will mean for people whose whole lives revolve around our jobs…that is, pretty much everybody. He is the the one who came up with the rather dystopian phrase “the gods and the useless” that Wardell seized on, and it is definitely a show stopper.
With that in mind, what does Chace say the impact of the economic singularity will look like?
Most people look at automation and think of a robot doing their job. That will eventually happen, probably before the technological singularity. But what will happen first is what is happening now: automation is slinking in and supplanting certain tasks, but not the whole job. In this way we are seeing partial automation of jobs that will, and is currently, putting people out of work while also devaluing their skills. That is, machines do not need to replace us completely, they just need to be better at our jobs than we are.
Currently, we see this everywhere:
- Robot assisted surgery, which makes surgery faster and more accurate, allows a single surgeon to do dozens more surgeries in a year than when forced to do them “by hand.” We’ve got more surgeries being done by fewer surgeons, and that ratio will continue to get bigger.
- Closer to home, the CVS drug store down the road from where I work has one human checkout station and five automated terminals.
- Banks have done similar with replacing bank tellers with ATMs. However, the claim is that bank tellers have grown in number, despite ATMs, is true. What is rarely mentioned is that bank tellers jobs have been revised from people who handle basic transactions to becoming de facto account managers, as well as the fact that banks opened more branches staffed with fewer tellers — both aspects of the job which has a limit on growth that has probably already been reached, or is close to maxing out.
- Amazon is putting big box chains out of business and people out of jobs, but also currently employs a lot of people for their storage and delivery services. Those jobs are already being “assisted” by robots and, eventually, that ratio of robots:humans will continue to rise. Soon, as in many manufacturing plants these days, a small number of people will oversee a robot army doing the hard-labor jobs.
- Financial planners are using algorithms and specialized software to do most of their jobs for them, likewise paralegals.
To see the long arc of this trend let’s go back to the crowd favorite: driverless cargo trucks. Right now truck driving is a reasonably well paid job that is done 100% by humans. At first, as trucks start to become automated, they will only be automated for specific tasks like highway driving. Well paid and experienced drivers will still be needed for heavy traffic zones such as cities and for parking (a complex and variable task). Next the trucks will advance to handling traffic better, but even then they will probably continue to have “drivers” on hand. However, those people won’t be driving, instead they will basically be tech support for when breakdowns occur or, perhaps, to handle parking in uniquely difficult situations. They will live in the truck’s cab, sleeping and reading and eating, as it travels without stopping between locations. They will eventually not be needed, but in the interim, some people will have those jobs. They will not be paid what highly trained and experienced truck drivers are paid, though, since that skill set won’t be needed. And of course, in the meantime, it will still exist as a job description. Until it doesn’t.
And that’s how the almost invisible switchover is happening with those jobs listed above, with the percentage of automation of a job rising under the guise of efficiency improvements, until the human is simply not needed anymore. We all know this is happening, to some degree. Chace puts it into perspective.
He also brings up convincing arguments about why this particular revolution is different from previous ones. It’s worth reading even if he doesn’t convince you, because he does bring a lot of facts to the fight.
One interesting thing he does, and does more often in current talks he gives than in this book, is look at the familiar argument that the industrial revolution displaced farm workers but found them new jobs in factories. Likewise, humans will still have jobs in the future despite automation because new jobs will be invented. Which is, after all, a comforting argument based on historical facts.
Chace pops that balloon by pointing out that we are not farm workers being shuffled to factories; we are the horses being outperformed by automobiles. Horses still exist, but they are not necessary for the economy to run anymore. We don’t need ‘em. No new jobs were invented to keep horses gainfully employed. They were retired (and likely sold off to be killed) and bred in fewer numbers. Nowadays only really committed horsefolk or rich people own horses. Who needs a horse? I don’t. You don’t. Publix doesn’t. Amazon doesn’t.
In this situation we are the horses, argues Chace. That should, I hope, give you pause.
Like Brynjolfsson and McAfee, he spends the last part of the book discussing our options. UBI is pulled out again, and surprisingly Chace does not think it is the solution, at least not in the short term. He thinks it is an eventuality, especially if/when the technological singularity happens, and that is what he means when he talks about the “death of capitalism.” (This man is no Marxist!) He believes that as things stand now, the world overall is not economically rich enough to pull UBI off, and it would cause as many problems as it solves. He’s surprisingly honest about the fact that he’s not entirely sure what the short term solution is, in fact, which I find refreshing. He ends by calling for governments to invest in exploring scenarios, through institutions such as think tanks and such.
He is, overall, optimistic. Given some of the scenarios he proposes, that’s a triumph in and of itself.
Conclusion
All three books hammer home this one point: do not judge tomorrow by the technology of today.
When new technologies appear, they slide in to take over technologies that have either maxed out or become burdensome in comparison. Early cars were terrible things, difficult to drive and dangerous, and most people assumed they would never be anything but curiosities. To this day we still use “horse power” to rate the power of a car’s engine even though automobiles have long since passed the actual strength and speed of even the strongest horse. It was hard for people to imagine, at first, riding in a “carriage” that could move faster than even the fastest horses could go, and so “horsepower” it was. Cars changed the standards if not the language.
For computers, that means that while the limitations of storage and speed can be calculated out to a definite limit, all the rules for those calculations change when new technology is introduced. That is exactly what happened when computers moved from vacuum tubes to transistors, and what will happen again when transistors give way to (possibly) quantum computers. Right now transistors are on track to reach the limit of miniaturization in 2021 (give or take, depending on how you quantify Moore’s Law), which will also define limits of capacity. But demands on transistors will increase, and so something will be developed and manufactured to meet those demands. Assuming that 20 years from now transistor technology will have stalled in 2021 at the point that their limits were reached is just plain naive.
Again, the evolution of autonomous cars is a good example of this thinking: a lot of people, even those familiar with concepts such as exponential growth and Moore’s Law, based their estimates of development on the technology at hand. If technological progress had frozen stiff at that point, then sure, their guesstimates would probably have held up. But of course it didn’t, technology exponentially grew and in a few years time driverless cars were being tested on public roads.
I mention this because throughout this post I have accepted as given that yes, both technological and economic singularities will happen. I haven’t seen any argument against that premise which isn’t based on current technological limitations. If all you are doing is looking around at what we have now and what it can do today, then you are not actually addressing the problem.
If your response is to cling to the belief that machines will never replace human labor, that there is something so uniquely special about what we do with our brains and bodies that no computer/robot could ever match it, then I hate to be the one to tell you this but: you are very wrong. You are wrong because you are thinking of what computers and machines can do today. You are wrong because you are looking at history and not towards the future.
The other argument against both singularity concepts is the refrain that the human brain is wonderfully complex and we don’t even know how it works, really, so it’s impossible to mimic. Which is true. The question “will computers ever mimic the human brain?” is the wrong question, actually. The human brain is pretty unique and amazing and so no, probably not (but…maybe someday?). In the short term the question is pointless because computers don’t need to mimic the human brain; they just need to mimic what we do with our brain. And that, like it or not, is not as complicated as we want to believe.
Machines will possess their own unique form of brain and intelligence, apart from ours, and will be able to mimic most of our jobs long before any computer can successfully mimic our brains. If the rate of computation speed and capacity continues to grow exponentially, then it really is possible that within a generation or two computers will be, for all intents and purposes, “sentient” (Kurzweil, ever the technology optimist, believes that computers will pass the Turing Test by 2029, while others think it will be 50-100 years from now), but even if it takes a long time to reach the technological singularity, the economic singularity will be crossed before then by a significant margin. That is honestly the most pressing concern.
We are living in uncharted waters concerning technology. Arguments may rage about timelines and where/when/how of specific developments, but those changes are coming. Kurzweil is unrepentantly optimistic about those changes, Brynjolfsson and McAfee are cautiously neutral, and Chace is pragmatic. I myself am borderline pessimistic.
If you get nothing else from those books (or this review of them) it needs to be this point: whether in five years or fifty, machines will replace professional, white collar, and blue collar workers — accountants, financial managers, teachers, nurses, lawyers, surgeons, truck drivers, manufacturing line workers, call center staff, novelists, middle management, and all service industry workers from janitors to wait staff. Even the most draconian sweatshops won’t be as cheap or efficient.
…what ya’ gonna do then?
* sorta kinda — from wikipedia: The scope of AI is disputed: as machines become increasingly capable, tasks considered as requiring “intelligence” are often removed from the definition, a phenomenon known as the AI effect, leading to the quip “AI is whatever hasn’t been done yet.”
** Aficionados of this topic like to make a big deal about the differences between “narrow AI” (non-sentient artificial intelligence that is focused on one narrow task) and “artificial general intelligence” (the intelligence of a machine that could successfully perform any intellectual task that a human being can) but again, it’s proving to be more of a spectrum shift than a tipping point from one to the other.
