Arnold Kling  

Ray Kurzweil's economics

Health Care Documentary... Housing and the Corruption Tax...

I discuss the economics of accelerating growth in my latest essay.

If output per person in 2025 is more than 5 times what it is today, then the economy will have won the race. That means that all of the concerns that economists raise about the middle of this century, such as the external debt of the U.S. economy (the cumulative trade deficit), the fiscal implications of Social Security and Medicare, or gloomy scenarios for global warming, will be trivialized by the sheer heights that economic wealth will have scaled by that time. If Kurzweil is correct, then the mountain of debt that we fear we are accumulating now will seem like a molehill by 2040.

See Ben Bernanke's speech on productivity for an overview of recent developments.

UPDATE: My second essay on Kurzweil is here.

For Discussion. Given the nonlinear nature of Kurzweil's forecast, how will we know that we are on the path to much higher growth until we are almost there?

Comments and Sharing

COMMENTS (24 to date)
Tom writes:


"To Kurzweil, this forecast would be ludicrously pessimistic. He would see it as an example of what he calls "intuitive linear" thinking, in which people forecast the future on the basis of a linear extrapolation of the past. For example, from 1960 through 1992, productivity growth in the nonfarm business sector averaged 1.6 percent. Accordingly, that may seem to be a reasonable rate of increase to project going forward."


"Kurzweil argues that the rate of technological innovation is doubling every decade, which to me would imply that the rate of productivity growth will double every decade. If annual productivity growth was 3.5 percent in the decade ending in 2005, then it will be 7 percent in the decade ending in 2015 and 14 percent in the decade ending in 2025. "

Isn't a doubling every decade also a linear trend? And isn't Kurzweil warning us about the dangers of linear thinking?!

Eli writes:

I'm skeptical. Clearly, technology begets technology, but I think that Kurzweil is overstating the rate of growth in technological innovation. It is, of course, very difficult to measure.

Even if Kurzweil is right that innovation has doubled every decade in recent decades, it is not at all clear (as Tom suggested) that this will continue. Some of it could be one-time gains from deregulation and tax cuts. While there is still a lot of progress to be made in these areas, they cannot go on forever.

Setting aside the business climate, it's still not clear that it's appropriate to extrapolate this trend. Perhaps a model of punctuated equilibrium would apply here: slow, steady growth in technological innovation until a catalytic technology is discovered which unleashes a flurry of innovation, which then slows down (and the process repeats).

Paul N writes:

No offense, but I find Kurzweil and the attention given to him embarrassing.

Part of his appeal must come from the fact that people would like it if his predictions came true, so they deceive themselves into believing them.

Mark Horn writes:

Tom says:

Isn't a doubling every decade also a linear trend?

No. Doubling every decade is an exponential trend.
10 linear steps: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
10 exponential steps: 1, 2, 4, 8, 16, 32, 64, 128, 256, 512

Eli says:

While there is still a lot of progress to be made in these areas, they cannot go on forever.
Why not?

I wish I could find it, but I remember reading an article about technological change and how it would soon peak. The article was written in the 1700's. History is littered with predictions that technological innovation "can't go on forever". They've been wrong historically and I believe that they'll continue to be wrong in perpetuity. Saying that progress can't go on forever is the equivalent of saying that we'll someday know everything that there is to know. This seems incredibly unlikely to me. So, while I can't prove that innovation can go on forever, that's where I'd place my bet.

Ivan Kirigin writes:

The _rate_ is doubling, not the state.

Decade | Increase that Decade | State
1 | 1 | 1
2 | 2 | 3
3 | 4 | 7
4 | 8 | 15
5 | 16 | 31
6 | 32 | 63
n | 2^(n-1) | 2^n - 1

Eli writes:

Mark, you quote me out of context (and perhaps I was not clear). I did not mean to say that technological growth cannot go on forever, but that you can only cut taxes and deregulate so much. Once the last regulation is eliminated, there are no more gains in productivity to be had from deregulation. Is that better?

Tom writes:
Isn't a doubling every decade also a linear trend?

OK, he did say the rate would double every decade, so it would be exponential growth. (Does he warn us about the dangers of exponential thinking?!)

If he has said labor productivity would grow at a fixed rate that also would be growing exponentially, right? But he said rate would even double every decade!

One other question to the mathematically inclined. Is the following statement correct then?

To Kurzweil, this forecast would be ludicrously pessimistic. He would see it as an example of what he calls "intuitive linear" thinking, in which people forecast the future on the basis of a linear extrapolation of the past. For example, from 1960 through 1992, productivity growth in the nonfarm business sector averaged 1.6 percent. Accordingly, that may seem to be a reasonable rate of increase to project going forward.

If I said productivity grew at 1.6 percent from 1960 to 1992 and I assumed the same rate of growth for the next 30 years would I be making a linear forcast or an exponential forcast? And would that be linear thinking or exponential thinking?

Barkley Rosser writes:

A doubling of the rate of technological innovations, which can be defended by looking at such things as the numbers of patents issued, does not necessarily imply any doubling of the rate of productivity improvement, if there are decreasing returns to R&D, which is not an unreasonable proposition. Indeed, we have seen this sort of acceleration of patents being issued for quite a few decades, but only minor changes from decade to decade on productivity improvement (did go up in the late 1990s, but seems to have gone back down a bit since to longer term averages since).

Dezakin writes:

I think this is misstating Kurzweil. 1% or 3% or 10% economic growth is all exponential, and so it looks linear unless you take longer time horizons.

But currently the rate of technological change is accellerating because of demographics. We are living in an era where around 3 billion people are slowly migrating from subsistance farming to industrial economies, and that brings more and more scientists and engineers to work on more specialized problems.

I do think Kurzweil is wrong about his dates, but not that wrong. But its fairly obvious that once you have human level AI traditional economics models go out the window. Your labor pool is limited only by your budget rather than how rapidly people reproduce.

Bernard Yomtov writes:

I agree with Barkley Rosser.

It's not at all clear that the rate of productivity growth must equal that of technological innovation.

My computer is many times more powerful than the one I had five years ago. I don't think my productivity on computer-based tasks has increased by the same factor, or anything close.

Stefan Karlsson writes:

This theory is just laughable. You're seriously embarrasing yourself by taking it seriously and even cautiosly embracing it. If it was some historical law that technological change must double every decade then how this didn't appear until the mid-1990s? And how come it didn't appear in other major economies?

And what evidence is there that this is the only thing driving productivity (who have recently slowed)? This productivity increase have in fact been more than enough cancelled out by slower growth in hours worked. While GDP growth was somewhat faster in the late 1990s than the previous decade, it has since fallen to a historically relatively low level of 2.5% during the 5 years of the Bush presidency.

Curt Gardner writes:

In the article, we find:

If annual productivity growth was 3.5 percent in the decade ending in 2005, then it will be 7 percent in the decade ending in 2015 and 14 percent in the decade ending in 2025. By that time, productivity would be more than 7 times what it is today. Thus, if average income per person is $35,000 today, then it will be over $250,000 per person (in today's purchasing power) in 2025.

My sense is that while technology innovation moves ahead dramatically in some areas, in other areas it doesn't have much impact. So, for example, in terms of buying computers, we do see a trend like the above if we consider the time period from 1970 to now. Back in 1970 it would have taken a fortune to buy the computer power that is today available for $1000.

But can we say the same thing about oil, for example, or basic medical care, or housing? Nowadays these things probably cost more today than in 1970 (in constant dollars).

Now maybe the technology innovation will power its way into more and more areas, forcing prices down (for example, maybe today's expensive medical procedures will become cheap). But I suspect there will always be some limiting factors, like waterfront land or genius or certain key resources, that will continue to demand a high premium. And as other things become cheap we devalue those things in our mind and continue to chase things that are scarce & have prestige.

Dave Milovich writes:

Kurzweil, as you've quoted him, is predicting economic growth will be doubly exponential (of the form a*exp(b*exp(c*time))). The Hanson paper you cite models world GDP as a finite series of singly exponential growth modes. Someone with more free time than I should take his data (which is included in the paper) and do a (nonlinear) regression with both doubly and singly exponential terms. Eyeballing Hanson's graphs, I'm guessing the doubly exponential terms will get neglible coefficients.

Dan Landau writes:

The weakness in Kurzweil’s approach is that it fastens on a particular technology as the key to faster growth, artificial intelligence. My father was talking about A.I. in the 1960s and it still isn’t here in 2005.

The real reasons why economic growth is will continue to accelerate are 3 fold. One, the technologies that will contribute to accelerating economic growth in the future are not limited to ICT. There is bio-technology, nano-technology, etc.. We don’t know which will lead for the next few decades and how far they can go. Two, the newly modernizing economies of China and India are soon going to flood the world with research scientists and engineers. How much that will accelerate the rate of scientific, technological, and productivity advance is impossible to say. However, it is sure it will accelerate the rate of progress. Three, more countries will be forced to modernize and free markets and protect property rights. This will happen because their old systems that slow innovation will leave them further and further behind the advancing free market economies. This why communism fell. The welfare state will disappear for similar reasons. Old Europe is falling further behind the Anglo Saxon countries. Eventually the peoples of old Europe will go for prosperity and creative destruction rather than security and stagnation.

technologydude writes:

1 - The limits of productivity growth are human, not technological. We are not increasing exponentially in our intelligence, nor in our motivation or our work ethic. Personally, I find myself less productive now in many ways than when I didnt have all the access to technology.

2 - As some become more productive, business gets more competitive, leaving many unemployable. The ratio of non-productive citizens to productive workers has been increasing, and this is a trend I see accelerating.

3 - Comparing productivity increases to cost increases in Medicare (or health care in general) is comparing something that is finite to demand, and cost, that is infinite. In our society, we generally do not even consider cost when it comes to keeping people alive.

Matt writes:

I share some of the skepticism, but to answer the question, if we see superfast growth, it probably won't come in all areas at once. Seeing very rapid progress in areas other than computing, perhaps in fields where the impact is less widespread and obvious, might be a sign that overall growth will eventually follow the optimistic projection.

Curt Gardner writes:

I just had a few more thoughts on this one... and wasn't sure I made my final point clear.

If we take the 1970-2005 comparison again, a key factor is that along with technological innovation we have rising standards and expectations. While from a 1970 perspective it would seem nearly unbelievable to get a 1995-era PC for $1K, from a 2005 perspective you'd essentially have to pay me to take that 1995 PC.

I expect the same thing to happen between now and 2025. By then we'll be awash with techno-baubles (things with built-in computing power), and the ones that are 5 years old will seem like trash to us (sort of like how today's iPods will seem worthless in a few years).

People's expectations continually shift as they get used to innovations, and what we consider a 'livable' technology level will rise. Both innovations and expectations seem locked in a ruthless cycle. (While I'm perfectly happy with a 1995 automobile, most Americans are not.)

In any case, I don't quite see how this technology innovation will magically take care of the trade deficit and various pledged entitlements... While perhaps we could in 2025 run a 'social security' program that would give people a 1990-level lifestyle, would that be considered acceptable to anyone?

And from the profit/wage point of view, my understanding is that real wages have been essentially stagnant in the US since the early seventies, and while corporate profits go up and down, I'm not sure there's any general trend toward profit. So where is the money supposed to come from to pay these future bills?

Dezakin writes:
The weakness in Kurzweil’s approach is that it fastens on a particular technology as the key to faster growth, artificial intelligence. My father was talking about A.I. in the 1960s and it still isn’t here in 2005.

A lot of 60's level AI is here now. Now I wasnt around during the time but all the books on the subject I've read were for doing things like automated theorem proving and expert systems; A lot of this stuff has made life easier in oblique ways.

Human level AI, where you download a human brain and then make x copies of it to slave for you (sidestepping potential ethical issues of reintroduction of slavery) requires hardware that wont be around for another two decades of growth in computer power for a reasonable sized group to do some experiments. Right now we're hard pressed to simulate nematode worms.

So we'll obviously develop brain simulation on inexpensive hardware sometime this century, the question is just weather its 30 years from now or 80.

Robert Schwartz writes:

I am very skeptical. I do not believe that trees grow to the sky. Nor do I believe that the most recent wave of technological advance is the most important that has ever occurred.

Enthusiasts like Kurzweil love to tout technology. But if we go back to the 19th century we can see technological revolutions that dwarf those of more recent years. Sure the internet is important, but is it nearly as revolutionary as the introduction of the telegraph? The steam ship and the railroad were a much bigger advance over the horse drawn carriage and the wooden sailing ship, than the 787 is over the 747.

Further, I would argue that every technology shows a similar growth pattern. It starts slowly, picks up speed, reaches altitude and then levels off. The process seems to take about 2 generations. the wright brothers first airplane flew in 1903. By 1968, Boeing was building the 747. Automobiles were first built in the 1880s. By the 1950s, they too had matured.

Both technologies have advanced but not at an exponential rate. Compared to the wright flier, the 747 was a different species. The 787 is an improvement over the 747, but it is not the a revolutionary advance.

Computer and semi-conductor technology will show the same pattern. Already we are seeing signs that the improvements will be slower in the future. Intel's abandonment of the P-IV was one. They couldn't make it go 4Ghz. I am willing to bet that .5nm process will be a long time coming and that we may never see .25. software has slowed visibly as it has gotten harder to do something revolutionary. Microsoft's inability to get the next Windows out the door is another sign.

What does all that mean in economic terms?

First, I would not look for accelerating growth.

Second, an important source of growth has been the integration of Russia, China and India into the world economy. This will continue to be important and beneficial to us for the foreseeable future. additional gains can come from Africa and the middle east.

Third, money does have a declining marginal utility. At some point, additional leisure will be more important to us than additional stuff.

Bottom line, I expect to see the American Economy grow for many more years, at rates similar to those we have experienced in the past.

AJ writes:

The technological possibilities are only an enabler of growth. The growth rate is the rate at which real people in orgainizations and jobs can make use and make changes in the way things are done and organize to do new products and services. This is a human, societal, business, legal, economic issue. Better organization and laws can increase it slightly, but it's a dynamic which accounts for the fact that economic well being has grown very consistently for hundreds of years in this country. There is something about humans and our way of life that has a potential economic growth of 2-5% - that's all we can handle (in the aggregate) in terms of rearranging the affairs needed.

Another example was the European discovery of America - the growth didn't happen overnight but was a major source of growth for 300 years.

So, you could dump the most amazing technological breakthroughs on society and we'd only have a slightly increased rate of growth in the economy for years.

It's this social/economic/political/business/legal arena which has a bigger effect upon the rate of growth and i don't see that changing much.

Roger D. McKinney writes:

Wow! Big meal to digest! Some quick observations:
1. McKinsey Co. found that the spurt in productivity in the late 1990's came mostly from Wal-Mart's model spreading throughout retailing. Productivity increases in services are much more difficult to achieve than in manufacturing and our economy is becoming much more service oriented.

2. The biggest threat to productivity comes from culture. Mariano Grondona writes ("A cultural Typology of Economic Development" in Culture Matters: How Values Shape Human Progress, ed. Lawrence Harrison and Samuel Hungtington) "If decisions favorable to development only answer to an instrumental value of an economic nature, such as increased wealth, the country's effort will decline as soon as the degree of wealth is attained." In other words, do we value the things that cause development (e.g., private property rights) for themselves, or because they make us wealthy? If we value them because they make us wealthy, we tend to abandon them once we achieve the expected wealth. I believe this is what we are seeing take place in Old Europe and what's beginning to happen in the US.

Thomas writes:

Tom says:

Isn't a doubling every decade also a linear trend?

but Mark Horn exactly aurgues that it's an esponential trend. The key point is: could one make such predictions for the futuure just basing his expectation on past values (both them having linear or esponential trend)? And then which past? Starting from when? It looks to me like the debate on adaptive or rational expectations. Moreover, Kling in his essay underline the fact that people understimate systematically the equity value on the long run because they forecast future on the basis of linear extrapolation of the past trend. But why the latter? Without a model on information and knolegde that explicate future expectations' process itself, this debate is poorly trivial.

mobile writes:

The economy may be growing non-linearly, but government has shown that it can keep pace. Even if we are all much richer by 2040, that mountain of debt could still look like a mountain of debt.

Dale writes:

Curt Gardner brings up an extremely important point. Even assuming Kurzweil is right, the growth won't be uniform. Those products for which we are most able to increase productivity will fall the most in price.

An interesting case to speculate on is art, music and literature because we are seeing a drastic disparity in the applicability of productivity gains in two areas. The technologies that have made artists, musicians and writers more productive are relatively minor. Certainly, there are writers who can write more productively with a computer than they could with a typewriter, but I suspect we are close to the limits of that gain. However, distribution is changing radically. The need to print and distribute physical copies is diminishing. Publication is shifting toward online distribution.

I think that we will see the price of creative activities, and of raw materials increase relative to the cost of products assembled from them. Those things that are not readily and cheaply reproducible will command the highest prices.

Comments for this entry have been closed
Return to top