this is an absolutely terrific critique of a disappointing book.
i'm an economist-consultant who's focused on long-term growth issue for 40+ years, know the history & data (us & global), & the literature (including acemoglu & his co-authors).
indeed, i read this book as soon as it was published. to my surprise, i found myself questioning many of the key arguments.
so i'm really glad to read this review: virtually every point seems to me clearly-stated, accurate, & relevant. taken altogether, usefully deflates a set of views that might otherwise be taken over-seriously.
Another strike against this book is its terrible scholarship: it frequently gets basic facts incorrect because the authors haven't bothered to actually research the topic. Example: it claims at several points that Eli Whitney was responsible for the development of interchangeable parts, a claim that has been widely and thoroughly debunked.
I am still sorting through all of this, but I had one quote I wanted to highlight.
"In Chapter 8, Acemoglu and Johnson blame communication technology for increasing inequality by enabling the offshoring of jobs to China and other countries."
Offshoring of jobs only increases inequality if you take a very narrow, very nationalistic view of equality. Inequality has only been rising within nations...between nations (as in all of humanity) it has been falling...why?? Globalization.
Millions of people have been lifted out of abject poverty and the same groups who would argue for borderless nations seem to get all bent that the decrease in inequality is coming for humanity as a whole...not their favored slice of humanity.
Does this mean inequality within a country is unimportant? Absolutely not. We need to minimize inequality if only to decrease the chances of demagogues destroying the goose that lays the golden egg. We should also reducing inequality within countries as because it creates more opportunities for more brilliant people to reach their full potential, which benefits all of us.
That said, I really wish people would stop conflating national increases in inequality with global increases in inequality. Especially brilliant economists like Acemoglu who should know better.
Western society has a long habit of solving problems ex post facto. In the Middle Ages, when it was discovered that emerging industries created noxious fumes, they were exiled outside medieval town walls.
In contrast, Acemoglu and Johnson want to substitute foresight for reaction: "their main idea is to redirect technological innovation toward technologies that complement workers instead of replacing them."
Again, turning to history one may note that the experience of Russia from 1917 onward does not speak well for the success of planned economies (Mandarins).
Indeed, who could have told at the time of the first cell phones how their capacities and roles would expand at the time of 5G?
Finally, appointing mandarins who would decide on the basis of limited information which emerging technologies were good for humanity would be a vast surrender of freedom. People are already resentful of elites who make decisions for them. Most people don't realize that the French Revolution was fought partly to gain economic freedom, both for entrepreneurs to start new businesses without government approval and consequently consumers to freely choose which products they care to spend their money on.
Kudos for your objective analysis of your friends work.
I honestly have not seen any wizards of technology "steer" anything. While I consider Gates and Jobs the two greatest entrepreneurs, business leaders, innovators with impact, they merely (in a grand way) could see/envision, the great unmet needs that Apple and PCs, amd software could serve.
Arguably one of the greatly (maybe changing) titans of business innovation, (act global, think local, be 1 or 2 in market share..etc) Jack Welch, was 100% clueless and disdained both Apple and Microsoft.
Welch Driven productivity and "work out". His efforts at driving this, were silently made possible by white collar productivity enabled by the PC. Gone were secretarys, admin and lower value add jobs. Draftsmen used to be (guestimate) 10:1 for design/mech engineers at GM. Now all engineers do their own design, most analyis.
As these tools developed from the 70s thru the 80s and early 90s, they had enormous impact.
Great review of a bad book. The idea that technological advances should be judged, taxed or subsidized based only on the impact of job reductions is not only impractical but counterproductive. If a drug that was a certain cure for various cancers was discovered, judging it by how many oncology doctors and nurses, administrators and morticians jobs are lost would be insane.
Wow... that was quite a complete review. Good work.. A couple of additional points....
1) Mental Model: The mental model in most economic analysis is an explicit separation of labor, investor, operator. In reality, the difference between human beings and entities (machines) is that humans simultaneously operate in all three simultaneously. Now, for any particular individual the spider chart of their magnitude of participation may change, but everyone does participate. Even the poorest person in the US is an "investor" by virtue of their "ownership" in the American enterprise, and their returns are the transfer payments from society.
2) Scope: In a worldwide integrated market, it is kind of funny to look at metrics such as income inequality in a local US scope. The fact of the matter is that income inequality fell significantly at a world-wide level... and yes. it was due to displacement of millions of American workers to the benefit of billions of Chinese/Indian/SE Asian workers. The other result of that process was a leap in productivity in the West.
I read tons of books, new and old, on prosperity and economic progress, but this one was one of very, very few that I rate as zero stars. I disagreed with almost every claim and every substantive paragraph in the entire book. It just came across as a sad attempt to use circular reasoning and questionable facts to get the reader to put people who think like them in charge. Yeah, that is “power.”
By the way, after reading your free substack for years, I finally agreed to subscribe. This is only my third subscription, and is based on both the quality and unbelievable quantity of your writing. Now I need to go back and read all the stuff I missed before!
I know that Daron Acemoglu is a legend in the Econ world and well respected by the field, but he also seems to get some fundamental basic things wrong. I have trouble squaring those two things in my head.
There is a movement within AI/Computer Science to focus on "human-centered AI". Some of the researchers in this movement share the motivations of the authors of this book. Others, like me, think that to create successful AI applications, the proper unit of analysis is the human+AI (or, more generally, the human organization + AI). Unfortunately, the AI, and especially AGI, mindset is all about creating artificial humans, whereas creating a successful product requires understanding the needs of the user and building a system to meet those needs. As a field, we have many metrics for measuring AI performance, but few metrics for measuring the robustness, resilience, and productivity of the human+AI system.
Your points about the difficulty of anticipating the effects of technology are very well taken. My experience as a participant in 4 startups is that the founders rarely know what the actual product and market will be, and startups must listen carefully to their early users and be ready to pivot quickly once they figure out what the real use cases are. If we consider LLMs, the industry is still trying to figure out what the LLM killer-app is. A smarter autocomplete is not going to be a huge productivity booster, for example. I'm much more optimistic about applications of deep learning in biology, pharma, and materials science than I am about LLMs in knowledge work.
I contest the idea that digital technologies led to higher inequality. Digital technology enabled the creation of global value chains and new outsourcing opportunities. Hence East Asia and European Europe industrialised and India got a services boom. This reduced global inequality. But all these lefty types will never bring that up.
My guess is that automation works differently if it's a non tradeable sector. The nursing example you gave shows how automation reduced labour. I'd imagine this happens because nursing services aren't tradeable and so automation doesn't increase output. What do you think of my assessment?
Do the authors talk at all about existential AI risk? If not, that's a more serious shortcoming than the ones identified here.
Noah's right to point out that since AI is a new and poorly understood technology, no one should make confident claims about how it will produce mass unemployment or what have you.
But I think the situation with existential risk is the opposite: just because it's a black box, nobody can be sure that superhuman AI *won't* bring about our extinction when someone carelessly assigns it a poorly specified task.
Yeah, what is going on with this book? The review and some of the comments are quite damning. I feel like there's a documentary to be made here on the writing of the book that will satisfy the curiosity of tens of people at least.
Having said that, there is a chance that AI and next gen robotics will be different because they will be able to automate so many different tasks, at least in theory. It'll be interesting to see what happens! I for one would be curious to hear more about wage subsidies and how those might work if automation really does take off. So far, we're basically only hearing about UBI as a possible policy remedy if AI automates much of the work we do now. I wonder what the full spectrum of policy options looks like.
“To sum up, footnotes and endnotes are a technology that has unambiguously benefitted the world, and even though they can be a bit of a pain the butt, authors should include them.”
Personally, I think this is even true of creative writing. It’s hardly possible to understand the meaning and cultural history an context of Dante’s “La Commedia” or Joyce’s “Finnegans Wake” without concurrently reading a companion volume of the complete annotations.
This is just spitballing, but I wonder if AI could someday generate accurate footnotes?
Cloud computing and bespoke chip design is putting Silicon Valley employees out of work at the highest tiers of workers. The same holds true at the lower tiers with companies reducing capex for in-house servers and personnel. Lower-tier workers, we learned during the pandemic, are, to a certain extent, essential workers. Ergo some upward pressure on wages. We do, after all, live in a physical, not virtual, world.
As for techno-optimists and Billionaire Bros: Who could forget that AAPL, GOOG, and MSFT were caught colluding to cap employees’ wages. And the lovely practice of back-dating besting of stock options.
It seems to me that AI and future technologies we can’t even imagine are, as they have been in the past, a function of the human factor: leaders and workers in tandem.
This is the type of article i subscribed for - thanks for a comprehensive review!
My one question is: can this be viewed as an attempt by economists to “integrate” themselves into direct policymaking basically? Who else would this “mandarins” be, if not economists? To me, this book seems like an application for a job that doesnt exist yet.
this is an absolutely terrific critique of a disappointing book.
i'm an economist-consultant who's focused on long-term growth issue for 40+ years, know the history & data (us & global), & the literature (including acemoglu & his co-authors).
indeed, i read this book as soon as it was published. to my surprise, i found myself questioning many of the key arguments.
so i'm really glad to read this review: virtually every point seems to me clearly-stated, accurate, & relevant. taken altogether, usefully deflates a set of views that might otherwise be taken over-seriously.
Another strike against this book is its terrible scholarship: it frequently gets basic facts incorrect because the authors haven't bothered to actually research the topic. Example: it claims at several points that Eli Whitney was responsible for the development of interchangeable parts, a claim that has been widely and thoroughly debunked.
I am still sorting through all of this, but I had one quote I wanted to highlight.
"In Chapter 8, Acemoglu and Johnson blame communication technology for increasing inequality by enabling the offshoring of jobs to China and other countries."
Offshoring of jobs only increases inequality if you take a very narrow, very nationalistic view of equality. Inequality has only been rising within nations...between nations (as in all of humanity) it has been falling...why?? Globalization.
Millions of people have been lifted out of abject poverty and the same groups who would argue for borderless nations seem to get all bent that the decrease in inequality is coming for humanity as a whole...not their favored slice of humanity.
Does this mean inequality within a country is unimportant? Absolutely not. We need to minimize inequality if only to decrease the chances of demagogues destroying the goose that lays the golden egg. We should also reducing inequality within countries as because it creates more opportunities for more brilliant people to reach their full potential, which benefits all of us.
That said, I really wish people would stop conflating national increases in inequality with global increases in inequality. Especially brilliant economists like Acemoglu who should know better.
Western society has a long habit of solving problems ex post facto. In the Middle Ages, when it was discovered that emerging industries created noxious fumes, they were exiled outside medieval town walls.
In contrast, Acemoglu and Johnson want to substitute foresight for reaction: "their main idea is to redirect technological innovation toward technologies that complement workers instead of replacing them."
Again, turning to history one may note that the experience of Russia from 1917 onward does not speak well for the success of planned economies (Mandarins).
Indeed, who could have told at the time of the first cell phones how their capacities and roles would expand at the time of 5G?
Finally, appointing mandarins who would decide on the basis of limited information which emerging technologies were good for humanity would be a vast surrender of freedom. People are already resentful of elites who make decisions for them. Most people don't realize that the French Revolution was fought partly to gain economic freedom, both for entrepreneurs to start new businesses without government approval and consequently consumers to freely choose which products they care to spend their money on.
Really nice review; I gave up on the book halfway through and feel vindicated.
Kudos for your objective analysis of your friends work.
I honestly have not seen any wizards of technology "steer" anything. While I consider Gates and Jobs the two greatest entrepreneurs, business leaders, innovators with impact, they merely (in a grand way) could see/envision, the great unmet needs that Apple and PCs, amd software could serve.
Arguably one of the greatly (maybe changing) titans of business innovation, (act global, think local, be 1 or 2 in market share..etc) Jack Welch, was 100% clueless and disdained both Apple and Microsoft.
Welch Driven productivity and "work out". His efforts at driving this, were silently made possible by white collar productivity enabled by the PC. Gone were secretarys, admin and lower value add jobs. Draftsmen used to be (guestimate) 10:1 for design/mech engineers at GM. Now all engineers do their own design, most analyis.
As these tools developed from the 70s thru the 80s and early 90s, they had enormous impact.
No one was the "lemming leader" technocrat.
Great review of a bad book. The idea that technological advances should be judged, taxed or subsidized based only on the impact of job reductions is not only impractical but counterproductive. If a drug that was a certain cure for various cancers was discovered, judging it by how many oncology doctors and nurses, administrators and morticians jobs are lost would be insane.
Wow... that was quite a complete review. Good work.. A couple of additional points....
1) Mental Model: The mental model in most economic analysis is an explicit separation of labor, investor, operator. In reality, the difference between human beings and entities (machines) is that humans simultaneously operate in all three simultaneously. Now, for any particular individual the spider chart of their magnitude of participation may change, but everyone does participate. Even the poorest person in the US is an "investor" by virtue of their "ownership" in the American enterprise, and their returns are the transfer payments from society.
2) Scope: In a worldwide integrated market, it is kind of funny to look at metrics such as income inequality in a local US scope. The fact of the matter is that income inequality fell significantly at a world-wide level... and yes. it was due to displacement of millions of American workers to the benefit of billions of Chinese/Indian/SE Asian workers. The other result of that process was a leap in productivity in the West.
I read tons of books, new and old, on prosperity and economic progress, but this one was one of very, very few that I rate as zero stars. I disagreed with almost every claim and every substantive paragraph in the entire book. It just came across as a sad attempt to use circular reasoning and questionable facts to get the reader to put people who think like them in charge. Yeah, that is “power.”
By the way, after reading your free substack for years, I finally agreed to subscribe. This is only my third subscription, and is based on both the quality and unbelievable quantity of your writing. Now I need to go back and read all the stuff I missed before!
Thanks Noah.
I know that Daron Acemoglu is a legend in the Econ world and well respected by the field, but he also seems to get some fundamental basic things wrong. I have trouble squaring those two things in my head.
There is a movement within AI/Computer Science to focus on "human-centered AI". Some of the researchers in this movement share the motivations of the authors of this book. Others, like me, think that to create successful AI applications, the proper unit of analysis is the human+AI (or, more generally, the human organization + AI). Unfortunately, the AI, and especially AGI, mindset is all about creating artificial humans, whereas creating a successful product requires understanding the needs of the user and building a system to meet those needs. As a field, we have many metrics for measuring AI performance, but few metrics for measuring the robustness, resilience, and productivity of the human+AI system.
Your points about the difficulty of anticipating the effects of technology are very well taken. My experience as a participant in 4 startups is that the founders rarely know what the actual product and market will be, and startups must listen carefully to their early users and be ready to pivot quickly once they figure out what the real use cases are. If we consider LLMs, the industry is still trying to figure out what the LLM killer-app is. A smarter autocomplete is not going to be a huge productivity booster, for example. I'm much more optimistic about applications of deep learning in biology, pharma, and materials science than I am about LLMs in knowledge work.
I contest the idea that digital technologies led to higher inequality. Digital technology enabled the creation of global value chains and new outsourcing opportunities. Hence East Asia and European Europe industrialised and India got a services boom. This reduced global inequality. But all these lefty types will never bring that up.
My guess is that automation works differently if it's a non tradeable sector. The nursing example you gave shows how automation reduced labour. I'd imagine this happens because nursing services aren't tradeable and so automation doesn't increase output. What do you think of my assessment?
Do the authors talk at all about existential AI risk? If not, that's a more serious shortcoming than the ones identified here.
Noah's right to point out that since AI is a new and poorly understood technology, no one should make confident claims about how it will produce mass unemployment or what have you.
But I think the situation with existential risk is the opposite: just because it's a black box, nobody can be sure that superhuman AI *won't* bring about our extinction when someone carelessly assigns it a poorly specified task.
Yeah, what is going on with this book? The review and some of the comments are quite damning. I feel like there's a documentary to be made here on the writing of the book that will satisfy the curiosity of tens of people at least.
Having said that, there is a chance that AI and next gen robotics will be different because they will be able to automate so many different tasks, at least in theory. It'll be interesting to see what happens! I for one would be curious to hear more about wage subsidies and how those might work if automation really does take off. So far, we're basically only hearing about UBI as a possible policy remedy if AI automates much of the work we do now. I wonder what the full spectrum of policy options looks like.
“To sum up, footnotes and endnotes are a technology that has unambiguously benefitted the world, and even though they can be a bit of a pain the butt, authors should include them.”
Personally, I think this is even true of creative writing. It’s hardly possible to understand the meaning and cultural history an context of Dante’s “La Commedia” or Joyce’s “Finnegans Wake” without concurrently reading a companion volume of the complete annotations.
This is just spitballing, but I wonder if AI could someday generate accurate footnotes?
Cloud computing and bespoke chip design is putting Silicon Valley employees out of work at the highest tiers of workers. The same holds true at the lower tiers with companies reducing capex for in-house servers and personnel. Lower-tier workers, we learned during the pandemic, are, to a certain extent, essential workers. Ergo some upward pressure on wages. We do, after all, live in a physical, not virtual, world.
As for techno-optimists and Billionaire Bros: Who could forget that AAPL, GOOG, and MSFT were caught colluding to cap employees’ wages. And the lovely practice of back-dating besting of stock options.
It seems to me that AI and future technologies we can’t even imagine are, as they have been in the past, a function of the human factor: leaders and workers in tandem.
This is the type of article i subscribed for - thanks for a comprehensive review!
My one question is: can this be viewed as an attempt by economists to “integrate” themselves into direct policymaking basically? Who else would this “mandarins” be, if not economists? To me, this book seems like an application for a job that doesnt exist yet.