You are here

Albert Wenger

Content Written by Author

Monday, June 18, 2018 - 12:53am

After nearly four weeks of not posting due to shoulder surgery I am almost back. I am saying almost because even though I can type very well again, I am spending a fair bit of time every day on physical therapy. That is time I would have spent writing and, well, something has got to give. So for now I am planning on one post per week instead of the usual three, but let’s see where it goes.

In the meantime though I want to thank everyone who kindly reached out, inquired how things were going, and wished me a speedy recovery. I appreciated every note! Hearing from friends and strangers helped a lot, especially in the early days post surgery when I was quite miserable. If you ever consider shoulder surgery for rotator cuff, just mentally prepare yourself for a really rough first week. You may wind up questioning whether it was a good idea, I certainly did. And while I am only four weeks out now I can feel improvements every day which makes me optimistic. 

If you care for a bit more background here is what happened. We went skiing in March to Verbier, Switzerland. On the first day in poor visibility I misjudged the distance from an off-piste run down to a cat track and there were about 5 feet of vertical. I hit the cat track hard, double ejected and pancaked into some very hard packed snow. I could hear a group of people standing near by go “Ouch.” I got up put my skis on an skied off but my left wrist and right shoulder definitely hurt. I skied the rest of the day but got fairly little sleep at night as my shoulder was throbbing. Given the great conditions though, I just took a lot of Advil the next morning and wound up skiing the entire five days we were there.

When I got back to New York my shoulder continued to hurt and I had fairly limited range of motion when trying to lift my right arm. I was hoping the whole thing would just go away with time but after six weeks without improvement I finally caved decided to see a doctor. Easier said then done because my insurance company wanted me to first get an x-ray before approving an MRI. The medical reasons for this are dubious at best and I decided to pay for the MRI out of pocket (you can get a much better rate when paying on the spot, one of the great distortions of our medical system).

With MRI in hand I went to two different doctors and got the same feedback: a mostly torn Supraspinatus tendon and a bunch of damage to the Glenoid labrum. Both required surgery to fix. The procedure itself was done at an outpatient facility and I was able to spend the night in my own bed (mostly not sleeping). Pro-tip: start taking the pain killers right away and don’t wait for the nerve block to wear off. Because of the nerve block you don’t feel any pain but when it wears off the pain is excruciating and pain killers take some time to work (file under: important note to self, should I have to do this again some time).

Now off to do my physical therapy exercises. And: Happy Father’s day to all fathers who made it to the end of this post!

Wednesday, May 23, 2018 - 11:30am

No Uncertainty Wednesday today. I had shoulder surgery yesterday to repair a rotator cuff injury. That means I won’t be able to type for quite a few days. I am creating this post using Voice on my Android phone. Impressively I did not have to correct a single word.

Monday, May 21, 2018 - 7:30am

NOTE: Today’s excerpt from World After Capital is about attention. It argues why attention is scarce in the sense of scarcity introduced earlier in the book. This section sets up the demands on attention and then talks about scarcity of attention for the individual (next time will look at scarcity of attention for society as a whole).


There is a limited amount of human attention in the world. We have 24 hours in the day and we need to spend some of that time eating and sleeping. For many people in the world much of their waking time is occupied by the job loop (both the earning and the spending parts). That leaves relatively little time for attention that we can freely allocate. This hard limit also exists in the aggregate, since—as I have argued earlier—we are headed for peak population.

At the same time that our attention is limited, we are using the Internet to dramatically increase the amount of available content. The increase in content is well documented to be exponential, which means that most of the content that has ever been produced by humanity has been produced in the last few years [47]. For example, YouTube alone is adding 100 hours of new video content every minute [48].

As a result, it is easy today to be completely overwhelmed by content. Our limited attention can readily be absorbed by ever refreshing content. Humans are maladapted to the information environment we now live in. Our brain evolved in a world where when you saw a cat, there was an actual cat. Now we live in a world of infinite cat pictures. This is analogous to our maladaptation to sugar for an environment that is now sugar rich (largely artificially so). Checking email, Twitter, Instagram, watching yet another YouTube clip or Snapchat story, or episode of one’s favorite show on a streaming service—these all provide quick “information hits” that trigger parts of our brain that evolved to be stimulated by novelty. As of 2017, the average person spends roughly two hours on social media every day [49].

The limited availability of attention has become the key new source of economic rents. Companies such as Google, Facebook and Twitter are valued in no small part based on the amount of attention they have been able to aggregate, some of which they then resell in the form of advertising. As a result they invest heavily in algorithms designed to present ever more captivating content to their end users in order to monopolize their attention. Sites like Buzzfeed and Huffington Post that are nominally news sites do the same.

Now even if you think this is problematic, does it mean attention is scarce in the precise meaning of scarcity that I defined earlier? That would require for us to not have enough attention to meet humanity’s basic needs. Is that really the case?

Individual Attention Scarcity

Let’s first consider attention at the individual level. All over the world people have constructed their identities around work and around firmly held core beliefs, whether religious or worldly. Both of these are undermined by digital technologies. We saw earlier how digital technology is putting pressure on labor. It is also putting pressure though on firmly held beliefs. Content is no longer easily contained in geographic boundaries and people are being exposed, many for the first time, to opinions and behaviors that diverge from their core beliefs.

In combination, this pressure is leading to a large scale crisis of individual identity and rising aggression both online and offline. This crisis takes many different forms, including increased teenage depression, growing adult suicide rates—particularly among middle-aged white males, and drug overdose deaths. In the US, these have increased almost 60 percent, 20 percent and 40 percent, respectively, between 2006 and 2015 [need more up-to-date statistics]:

This is not dissimilar from the beginning of the Industrial Age, when people had to leave the countryside and move to big cities. They were forced to give up identities that had been constructed around land and a historical set of professions. They were confronted with people from other regions who held different beliefs.

Just as with the transition into the Industrial Age it is therefore not surprising that there is a rise in populist leaders with simplistic messages, such as Donald Trump in the United States and Viktor Orban in Hungary. A recent study found that throughout Europe, populist parties are receiving more than double their average share of the vote in national and parliamentary elections compared with the 1960s [50]. People whose identity is shaken want to be reassured. They want to hear that things will be OK and that the way of getting there is simple. “Make America Great Again” is an example of that. So is ISIS. In both cases the message is retrograde. Instead of a new identity that has to be built, requiring time and effort, these backward movements promise an easy return to a glorious identity of the past.

Our attention to our most basic need, the existential need to make sense of the world as an individual by finding a purpose that makes our life meaningful, is scarce. Instead we let our attention be occupied by our job or by yet another video or worse by propaganda. This individual scarcity of attention is not confined to any one demographic. Definitely people who have to work multiple jobs just to make rent and feed their families are impacted. But so are many people in high paying jobs who are often working more hours today than they ever have.

I do a fair bit of counseling for young people who want to work for a technology startup or who want to enter venture capital. Most of them are looking for tactical advice, such as how to apply to a specific position. After discussing that for some time, I usually switch gears and ask them a much more open question. “What do you want from your next position?” That often elicits answers such as learning a new skill, or applying a skill that they have recently learned. Sometimes people answer with a desire to contribute to some cause. I then get to the point by asking directly “What is your purpose?” Shockingly few people have an answer to that.

Purpose is an individual need for which the Industrial Age had little use. Somebody with a strong sense of purpose does not fit readily into the job loop either as a worker or as a consumer. Instead work and consumption have become the de facto purpose for most people. Both the cultural and religious narratives adjusted from the Agrarian Age to the Industrial age to support this re-definition of purpose.

With digital technology we can now exit the job loop and redirect attention to finding other sources of purpose. Instead though we are using digital technology to aggregate attention primarily for resale (advertising) and for entertainment. We do not identify this as a fundamental problem of the largest platforms, focusing instead on areas such as privacy and moderation of speech. That’s because we continue to see the world through the lens of capital scarcity instead of attention scarcity.

Wednesday, May 16, 2018 - 5:10pm

Now that we have spent the last few Uncertainty Wednesdays on modeling beliefs as probability distributions, we can now get to the topic of updating. Updating is what we are supposed to do with our beliefs when we have new observations. We first encountered a similar idea in the extensive example of a cancer test which we used to derive Bayes’ theorem.

In that post I wrote that “[Bayes’ theorem] relates the probability of the world being in state B *before* we have observed a signal to the probability *after* we have observed signal H.” Now in that quote and in the example we used probabilities and not probability distributions. We had found the following formula, which is known as Bayes’ rule:

P(B | H) = [P(H | B) / P(H)] * P(B)

As a reminder P(B) is the probability of event B before we have observed anything, so in the case of cancer this would be the rate at which this cancer occurs in the relevant population. P(B | H) is the updated probability conditional on having received a positive cancer test (H as in high). We saw that the updating occurs through the factor P(H | B) / P(H) which consists of the sensitivity of the test P(H | B) divided by the total probability of seeing a positive test P(H).

What we are looking for now is to come up with a similar version of Bayes’ rule for beliefs expressed as probability distributions. We want something that looks roughly like:

posterior belief = update factor * prior belief

where again the update factor captures the likelihood of the observations, keeping in mind here that we are now dealing with distributions. The beliefs and the update factor are both functions which makes the formula for this quite daunting looking. We will write it – with some abuse of notation – as follows to keep things simple 

p(θ | x) = [p(x | θ) / p(x) ] * p(θ)

where θ is the parameter we are interested in, such as the probability of Heads for our coin, and x denotes our observations. We see the numerator of the update factor now is p(x | θ) – this is a function, the so-called likelihood function, which maps θ into p(x | θ). The denominator is p(x) which is the probability of the observations. That in turn is quite complicated if we unpack it, since it is an integral over all the possible values of θ and their probabilities (what comes out though is a scalar, meaning just a number, not a function).

So what we are really doing is multiplying two functions: the likelihood function and the probability density function of our prior belief which gives us a new function that represents our updated or posterior belief. This is complicated for the general case and along with calculating p(x) by evaluating the integral will require numerical approximations. 

Thankfully though it turns out that there are elegant and simple solutions for some types of probability distributions, such as the Beta distribution which I had introduced as an example of a possible belief. If you have a Beta distribution as the prior belief for the probability parameter in a coin toss then the posterior belief also takes the form of a Beta distribution! Next Wednesday we will use this fact to show how we can easily update our beliefs on a coin toss (provided we buy into using the Beta as the shape of our prior). 

Monday, May 14, 2018 - 5:05pm

NOTE: Today’s excerpt from World After Capital is about labor. In it I describe what I call the “job loop” and how it has become central to both the economy and sadly also our view of human dignity.


Before we can get to attention, though, we need to discuss the changing role of labor in the economy. Thinking about labor is hard because of an odd interweaving of cultural beliefs with economic history that I will try to disentangle. Over the last couple hundred years we have convinced ourselves that employment is essential both for the functioning of the economy and for individual dignity.

Let’s start from the perspective of production. If you want to make products or deliver a service you require a series of inputs, including buildings and machines (capital), raw materials or parts (supplies) and, historically, human workers (labor). For much of history, capital and labor turned out to be complements. As the owner of a company you really couldn’t make use of the company’s physical capital without having labor to operate it. That was true for manufacturing and holds even more so for services, which often use very little capital and consist primarily of labor.

But, and this is where it gets confusing, there is nothing in economics that says any particular production process has to require labor. The deemed necessity of labor happens to be an artifact of the production functions that were technologically available to us when economists started to develop the theory of production. If you, as the owner of a company, figure out through technological progress how to do something with less labor, or no labor at all, and that form of production is cheaper than before, that’s what you will choose to do. WhatsApp, when it was acquired by Facebook for $19 Billion had fewer than 50 employees!

There would seem to be a catch though. While having no labor might make sense for any one company, for the economy as a whole, who is going to buy all these goods and services if people are out of work and hence don’t have any money? There is the famous story about an exchange between Henry Ford II and Walter Reuther who then headed up the Automobile workers union, which went as follows:

Henry Ford II: Walter, how are you going to get those robots to pay your union dues? Walter Reuther: Henry, how are you going to get them to buy your cars? [42]

Now if we all had inherited wealth, or sufficient income from capital, an economy without labor wouldn’t pose a problem. As a first approximation, we would have the same demand. But none of us would have to work and all of us could enjoy the benefits of cheaper products and services courtesy of robots and automation.

The Job Loop

For a long time the possibility of a slump in consumer demand due to less labor seemed not just unlikely, but downright impossible. We had a perfectly working loop at the heart of economic growth, which I call the “job loop”

In today’s economy the majority of people have a job. They sell their labor, producing goods and services for someone else and receiving wages in return. They then take those wages and go buy stuff. Smart phones. Books. Tools. Houses. Cars. Gas for their cars. They also buy services, the professional assistance of attorneys and doctors and auto-mechanics and gardeners and hair stylists and nutritionists. Most of the people who sell them goods and services, in turn, are employed and take what they are paid and live on that, buying goods and services from still other people.

The job loop worked incredibly well in combination with competitive markets for goods and services and a properly functioning banking and finance system. Entrepreneurs would come up with new and improved offerings. They would use debt and/or equity to start new businesses which would employ people (often at higher wages than older businesses, giving employees more purchasing power). It was an amazing virtuous cycle that resulted in unprecedented prosperity and innovation.

A quick aside, as some might say that many people these days are self-employed or independent contractors. For the purposes of this analysis that is irrelevant as long as they are fundamentally selling their time. For instance, a graphic designer who works as an independent contractor (freelancer) is still largely paid for the time they put into a project. It is only if the designer can develop something, say a graphics template, that is paid for over and over without further time spent that they exit the job loop.

The problem with any virtuous cycle is that the effect of mutual re-enforcement applies just as much in the other direction when things contract. Take a small town, for example, in which local stores provide some of the employment. Now a big superstore comes into town, resulting in reduced total retail employment and lower wages. Yes, maybe the products they sell are cheaper also, but it is entirely possible to set off a contractionary cycle. Fewer store employees have income (and those who do have less). They start spending less on haircuts and car repairs. That means the hair dresser and car repair person earn less and can spend less at the local restaurant.

Could this happen to the economy as a whole?

The Great Decoupling

We are in the middle of a version of that playing itself out at the scale of the U.S. economy and potentially the global economy as well. It starts with what has become known as the great decoupling. For a long time as the economy grew, the share of GDP going to labor grew right along. Beginning around 1980 though GDP continued to grow while household income remained flat — hence the term “decoupling”

But GDP continued to grow so what’s the problem? Well much of that growth was financed by consumers going into debt instead.

Eventually there was a limit to how much debt households could support. Once we reached that limit, we had the making of a situation of insufficient aggregate demand in the economy. The first event that really drove that point home was the collapse of the U.S. housing bubble. There is a fair bit of evidence that we are hitting another such point right now. In theory, the dramatic decline in oil prices should put more money into the hands of consumers and stimulate demand. Instead, it appears that consumers are using it to pay off debt and even start to save.

What’s driving the decoupling? Part of it may be demographics, but part of it is likely to be the impact of technology. To the extent that accelerates, as I believe it will, there will be further pressure on aggregate demand. From a traditional economic growth perspective what should be particularly worrisome is that jobs in developing countries have a high exposure to automation [43]. Put differently, these countries may skip the golden age of the job loop entirely or have a much diminished version.

We don’t need an indefinite growth of aggregate demand to take care of basic needs (wants by contrast are unlimited). Nonetheless, a rapid demand collapse would be a bad thing for societies that, for now, are built largely on the job loop. That raises the question of whether it is at all possible for technology to depress wages over a prolonged period of time.

Lump of Labor or Magic Employment Fallacy?

With the job loop dominant, people have to sell their labor to earn a living. Until recently most economists didn’t worry at all about this ever being an issue. They believed that when human labor gets replaced in one part of the economy, say agriculture, it finds work in another part, say manufacturing. When manufacturing starts to get automated, labor is sought after for services. These economists refer to a fear of technological un- or under-employment as the “Lump of Labor Fallacy.”

The argument goes something like this. We automate some part of the economy. That frees labor up to work on something else. Entrepreneurs use this newly available labor to deliver innovative new products and services. There is no fixed “lump” of labor, rather there are potentially infinitely many things to work on. As support for their argument they offer that this is exactly what has happened historically. And so they ask, why should this time be different?

To understand how things could be different, it is instructive to consider horses. As recently as 1915 we employed over 25 million horses in the U.S. in agriculture and for transportation. By 1960 that number had declined to 3 million and then we stopped keeping track systematically [44]. What happened? Well, we figured out how to build tractors, cars, and tanks. There were no use cases left in which horses were superior to a mechanical substitute. The potential for the same to happen to humans was pointed out by economist Wassily Leontief in his 1952 work, Machines and Man [45].

Some people will immediately object that, well, horses can’t think and obviously we humans can, giving us a far broader range of things to do. That is true and is also the reason why so far we have always found new employment for people. So what has changed? Well, as we saw in the chapter on Digital Technology, we now have computers and we have figured out how to have computers do lots of things that until quite recently we thought only humans could, such as driving a car.

With digital technologies we have universal machines at zero marginal cost. All of the sudden the idea that we might be like horses, and have fewer and fewer uses, doesn’t seem quite so impossible.

Those who continue to claim this is committing the “Lump of Labor Fallacy” immediately retort that this simply signals a lack of imagination. They argue that we just haven’t thought of some new set of human activities that will once again gainfully employ people. But that line of thinking could also be a fallacy. I will call it the “Magic Employment Fallacy.” Just because we have found new employment in the past, doesn’t mean we will in the future, especially when we have an entirely new set of technological capabilities.

Yes we humans can be incredibly creative and think of new things to spend our time on. But the operative question for people selling their labor is not if they can think of something to do, but if they can get paid for it. Not just get paid something, but enough to cover all of one’s basic needs. It doesn’t matter what creative pursuit or new service we think of, the only thing that matters is whether a machine (or another human for that matter) is capable of doing it more cheaply.

This, in particular, turns out to be a problem with the “Magic Employment Fallacy.” Nothing in economics says what the clearing price for labor ought to be (the wage level at which there is no unemployment, and no shortage of labor). It could happen to be well below what people need to cover their basic needs. And that means we have a problem even if everyone were employed, unless you want to make an argument that we should simply let that happen and eventually wind up with fewer people, just as we did with fewer horses.

So in order for the “Magic Employment Fallacy” to, well, not be a fallacy, we have to find new high value things for humans to do for which there is both paid demand and machines are not effective substitutes. I don’t think we can rule that out entirely. We may find that the best candidate is a cultural shift that leads us to value goods and services produced by humans qua human production. The success of marketplaces, such as Etsy, that sell handmade goods, and the rise of artisinal goods more generally, are potential indicators of such a shift.

Another area where we may value humans qua their being human is in caring for the young, the elderly and the sick. Given changes in demographics we will need significantly more care for the elderly. Yet while we may want to value human care more highly, there is a potential wealth distribution issue. For instance, many people in the U.S. (and elsewhere) don’t have the savings that would allow them to pay for human help as they get old.

Whether it is Lump of Labor or Magic Employment, at a minimum we have to be prepared for a potentially long adjustment period. That alone is an argument for a need for increased economic freedom (see the later chapter) but there is a much more powerful one: we should prefer automation over human employment.

Expensive Labor and Innovation

Some people argue that unions were bad because they made labor expensive, which resulted in costly products and services that people could not afford. There is, however, a completely different way to look at unions raising the price of labor: it propelled us to become more efficient by creating a problem that entrepreneurs had to overcome, and the way they overcame it was through innovation—by building better machines that required fewer humans. One can still see the negative effects of abundant cheap labor in places such as India. There is little incentive to invest in a machine, if it is cheaper to have people do the work by hand.

It is bad to be stuck in a low innovation trap. Now we face this risk globally. The combination of a fear of automation and some automation making labor cheap could have exactly that effect. For example, we could easily wind up with many more years of people having to drive trucks back and forth across the country, long after a machine could do the same job and do it safer [46]. Pick any other job, say toilet cleaning, and ask what the incentive is to automate that job as long as you can get someone to do it for minimum wage?

Some will object to automation on a totally different ground though. They will argue that people require work as an integral part of their identity. That work is what gives humans dignity. If you have been a truck driver for a decade or more, who are you, if you can no longer earn a living driving a truck? This is an area where unions have historically been problematic: trying to preserve jobs for the sake of carrying out that job and also to preserve the union itself, which represents those employees. Today though it may just as easily be politicians who proclaim that jobs must be protected as a source of dignity.

So now we see what we need to solve for with regard to labor: a way to embrace automation without a collapse in aggregate demand, while simultaneously getting away from the idea that a job is the source of human dignity. This may seem like an outrageous claim to some and is certainly a tall order. But the next section on the scarcity of attention will explain why it is critically important.

Wednesday, May 9, 2018 - 5:10pm

Today’s Uncertainty Wednesday is a quick wrap up to the formalization of beliefs as probability distributions. Let’s first start with a simple question. What probability distribution reflects the least knowledge about a coin? How do you express “I don’t know, the coin could be anything from only heads to only tails to anything in between”? The answer is a uniform distribution as shown here:


The horizontal axis is the probability of the coin coming up heads – ranging from 0 to 1 and the vertical axis is what exactly? Well this depends on whether you paid attention to last week’s explanation of probability density functions (pdf) versus cumulative distribution functions (cdf). Which one is this?

Well it is a pdf, to be precise it is the pdf a uniform distribution over the interval [0, 1], which means that the vertical axis helps us measure the probability over any (sub)interval of the horizontal axis. The area under the curve is the probability. We can easily verify that the total is 1 because 1 x 1 = 1. Now consider the probability between say 0.4 and 0.6. That’s (0.6 - 0.4) x 1 = 0.2. So our belief here is that P(H) (probability of observing Heads) will be in the interval between 0.4 and 0.6 with probability 0.2.

So between last Wednesday and today we have seen how to express a range of beliefs from knowing nothing to certainty as probability distributions. Starting next Wednesday we will get to the question of how to represent the updating of beliefs when we make new observations.

Tuesday, May 8, 2018 - 7:30am

NOTE: Today’s excerpt from my book World After Capital deals with how we have achieved the sufficiency of physical capital. This is capitalisms greatest accomplishment, but also means that we are now facing a new scarcity: attention.


The title of the book is World after Capital. One of my fundamental claims is that capital is no longer scarce. There is enough capital in the world to meet everyone’s basic needs. That means meeting the individual needs of 7 billion or more people, the collective needs of the societies they live in and the collective needs of humanity at large. Using the language introduced earlier, capital is sufficient. And because population growth is decelerating, while technological progress is accelerating (due to digital technology), capital will no longer be the binding constraint for humanity going forward.

It is tempting to look at this in terms of financial capital, but that again would be succumbing to the veil of of money, as was the case with the definition of scarcity. Dollar bills don’t feed people. Gold bars can’t be used as smart phones. The capital that matters is productive physical capital, such as machines and buildings.

Financial capital is not irrelevant. It is generally required both for the initial construction of physical capital and to meet the ongoing working capital needs of the economy. If I want to build a factory or a school, I need to pay the construction workers, the suppliers of machines, etc. before I can start collecting money. And in many businesses I pay some ongoing expenses every month before collecting revenues from customers. Cash outflows preceding cash inflows means a financing mechanism is required. To get the proper accumulation of physical capital, we therefore need to have effective ways of accumulating and allocating financial capital.

In the history of financial capital there have been many important innovations, such as corporations with limited liability, debt and equity issuance and trading, bank lending and more recently market place lending. The allocation of financial capital to projects through markets has been enormously successful, compared to attempts at various forms of centralized planning. It is the very success of the market-based approach that has now given us a physical capital base in the world that is large enough to meet our basic needs.

Many recent innovations in finance, however, have not contributed meaningfully to the proper creation and allocation of physical capital. Quite the opposite. They have contributed to the “financialization” of the economy: a growth in financial sector activities that is decoupled from or even harms the formation of physical capital. For instance, many derivatives and structured securities have resulted in severe misallocations by shifting risk. One example is the housing bubble that resulted in part as mortgage backed securities and CDOs appeared to remove all risk from capital flooding into construction.

What is the role of “human capital” in all of this? Human capital is the subset of all knowledge that embodied in a group of humans. So the question is better asked differently: what is the role of knowledge? The answer is that advances in knowledge are essential for making capital more effective. Even more fundamentally, knowledge is necessary for having physical capital in the first place.

You can theoretically have physical capital without financial capital but you cannot have physical capital without knowledge. You cannot build a machine, say an MRI, without a lot of knowledge in physics and engineering. In a world where everyone’s basic needs are taken care of it might, however, be possible to build the same MRI without the need for financial capital.

Interestingly, you can also have financial capital without physical capital and without meaningful knowledge accumulation. For instance, you can develop financial capital through trade or war or simply by convention as in the case of the island of Yap [40].

All of this is to say that we should never lose sight of the fact that financial capital ultimately serves no purpose in and of itself, other than possibly the gratification of ego. As great illustration of that imagine a Spanish Galleon full of raided gold sinking in a storm. The sailors aboard had ample access to financial capital, but what they really needed to survive was more knowledge and better physical capital.

So now we will go ahead and examine whether physical capital is still a binding constraint when it comes to meeting basic needs. The approach I am taking is split in two parts: here in the main text I am applying logic based on observations; the Appendix contains much more data and calculations to back up the arguments.

Individual Needs

My claim is that capital is no longer the binding constraint for meeting individual needs, not just for one individual but for everyone. This is especially true for the developed economies but increasingly true globally.

The primary strategies for meeting our power needs are breathing air, drinking water and eating farmed food.

There is plenty of air to breathe (one time reminder: please see the Appendix for backup on this and the following assertions), the key challenge today is having clean, breathable air. China and India are both struggling with that at the moment, but this is due to rapid development using outdated energy sources. The clean air achieved in industrialized countries shows that this is a temporary development stage.

Similarly there is plenty of water in the world for everyone to drink. There are distribution and access problems, including right here in the United States (e.g., the polluted water in Flint, Michigan). Again though, physical capital is not a binding constraint. We can even build new desalination plants in record time. [Example]

We have also made dramatic progress in farming. In fact, globally the amount of land required for farming has started to decline as a result of higher per acre productivity. We have made recent breakthroughs in vertical and automated farming. For instance, the world’s largest vertical farm is currently under construction in Jersey City. The Japanese indoor farming company Spread is working on a fully automated facility that will be able to produce 30,000 heads of lettuce per day [155].

The discharge need is primarily addressed through modern sewage technology. Here too capital is no longer a binding constraint per se, but again there is a global distribution problem. To see how quickly this has the potential to change, consider the migration that has taken place in China from the country side into cities.

The Chinese construction boom also illustrates how quickly we can build shelter as a strategy to address the need for a controlled physical environment. In the U.S. too we had a prior construction boom which was powered by artificially cheap mortgage credit. While a lot of housing was built in the wrong places it powerfully demonstrated our construction capacity.

Clothing is another strategy for addressing this need. The price of clothing has been falling in the United States and in many other parts of the world. Capital is not a constraint here and we can clothe everyone in the world many times over.

Similarly we have become very good at providing light. There is a great study that shows how the hours of light one can earn with 60 hours of labor have exploded in the United States from about 10 in 1800 to over 100,000 by 1990 [CITATION?]. We have made further progress since with LED lighting. That progress has also come to other parts of the world, for instance in the form of off grid solar powered lamps.

Now we come to a more difficult need, the one for healing. We read all the time how expensive healthcare has become and how it consumes an ever larger fraction of the economy, at least here in the United States. We have to ask though whether capital really is a binding constraint here. Again in industrialized countries this does no longer appear to be the case. We have plenty of hospital space and doctor’s offices. We have extensive diagnostic facilities and can produce large quantities of medicine. The binding constraint instead is one of insufficient knowledge. Our bodies are extremely complex and even seemingly basic issues, such as how diet relates to health, are poorly understood as a result.

In learning we are also no longer capital constrained. This is rapidly true not just in industrialized nations but also globally due to the buildout of wireless networks and the increasing affordability of smartphones. We are not far away from a point in time when we have enough capital for anyone in the world to learn anything. The binding constraint here is not capital but the availability of affordable content and the time to learn (and to teach).

The final individual need, the one for meaning, is not and has never been constrained by capital.

Collective Needs

At first it might seem difficult to see how capital even relates to our collective needs as defined in the earlier chapter. How could capital have anything to do with such abstract concepts as motivation and coordination? Was capital ever a binding constraint here?

Capital clearly was not a binding constraint for reproduction, which societies thankfully accomplished a long time ago or we would not be here today.

But when it comes to allocation, capital was the crucial binding constraint during the Industrial Age. Not only were we terribly bad at making stuff at first but we also lacked the communications and transportation infrastructure to easily get goods to where they were needed.

Motivation might historically appear not to to be capital constrained as we had many strategies for the motivation need, including rewards and punishments. The development of markets with prices, however, turned out to be a crucial strategy for meeting the motivation need. High prices provide an incentive for the allocation of capital (and other factors of production). For a long time capital in turn was the binding constraint on the scale of markets. Today, however, we can broadcast supply, demand, and prices in any market globally in near realtime at zero marginal cost.

Coordination, on the other hand, was quite obviously capital constrained for a long time due to limitations on communications. We can see this by considering that until fairly recently it was not possible to have a globally coordinated event. Today on the other hand we not only have a global nearly instantaneous communication network but also the ability to precisely position people or machines using GPS and other location services.

Finally, our collective need for knowledge was capital constrained for a long time. Making books for instance was expensive and time consuming. Copies of books had to be made by humans introducing errors. The spread of knowledge was constrained by the need to create and move physical copies. We have now left all of those capital constraints on knowledge behind.


Our progress on enablers is another way to understand why capital is no longer the binding constraint. We have had massive breakthroughs on all four during the Industrial Age: energy, resources, transformation, and transportation.

The biggest breakthrough in energy was the development of electricity. It allowed us to apply energy in highly precise fashion. Our remaining challenges are all related to the production, storage and distribution of electricity. Further improvements in energy will let us solve needs in new ways, but we are not fundamentally energy constrained today. For instance, a relatively small percentage of surface coverage with solar (< 1% in the US) would cover all electricity needs at current efficiency rates [SOURCE?].

Resources were also completely transformed during the Industrial Age through mining, which in turn was enabled by progress with transportation (rail) and energy (steam power). People, especially those motivated by a concern for sustainability, like to point to scarcity of resources as the primary constraint. But resources are sufficient when we consider three sources that we can tap in the future: recycling, asteroid mining and transmutation. For instance, today a lot of electronics wind up in landfill instead of the materials being recycled. We achieved the first soft landing on an asteroid as far back as 2001. And while transmutation sounds like modern day alchemy, we now routinely make phosphorus out of silicon (albeit in small amounts).

Our ability to transform also improved radically during the Industrial Age. For instance, chemistry allowed us to make rubber synthetically which previously had to be harvested from trees. With machine tools, such as drills and lathes, we were able to rapidly transform wood and metals. Later we added transformation technologies such as injection molding and more recently various additive manufacturing technologies (often referred to as 3D printing).

Transportation went from human powered to machine powered dramatically changing our capabilities. We went from walking to traveling to space in rockets. We can fly across continents and oceans on commercial flights and reach any major city by air in just a day (or two at most). While some have complained about a lack of progress in flight, pointing to the lack of commercial supersonic options following the retirement of the Concorde, we had extraordinary progress in flight safety. More recently work has resumed on new options for commercial supersonic flight and we have made tremendous progress with reusable rockets and closer to earth with autonomous vehicles (for instance drones and warehouse robots).

The progress on these enablers has allowed us to produce more physical capital, do so more rapidly and cheaply, and transport it to anywhere in the world. One way to appreciate just how far we have come is to note that the first time smartphones became available was only in 2000. By 2017 over 8 billion smartphones had been produced and shipped and there are currently over 2 billion smartphone users in the world.

As an important reminder before moving on. I am not claiming that everyone’s basic needs are being met today. Far from it. Nor am I arguing that governments should be using central planning or that they should be meeting people’s basic needs through government run programs such as food stamps or subsidized housing (in fact quite the opposite, as I will argue later when writing about economic freedom).

The point of this chapter is simply to argue that physical capital is no longer the constraint in meeting everyone’s basic needs. We are not dealing with a problem of capital scarcity—in the sense of technological scarcity introduced earlier—but with one of allocation and distribution.

Capital is no longer scarce but sufficient. We should consider that the great success of capitalism.

We now face a new scarcity, however, that of attention, and capitalism will not solve it for us without changes in regulation and in self-regulation. Before we can examine the scarcity of attention though we need to understand how digital technologies have the potential to change the role of labor.

Friday, May 4, 2018 - 11:30am

I recently finished Nassim Taleb’s latest book “Skin in the Game.” Much like Antifragile previously, I highly recommend reading it. The subtitle of the book is “Hidden Asymmetries in Daily Life” and while I am not a fan of the obsession of publishers to add a subtitle to every non-fiction book (try finding one without), asymmetries are the leitmotif that runs throughout, including a wonderfully succinct table on “Asymmetries in Society.”

The book is in the classical philosophical tradition of lessons that one can actually apply to one’s own life, ranging from how to pick a surgeon to how to maximize one’s freedom to act. Taleb is at his best when he ties together mathematical analysis with observations about present day society and then relates it all back to a long history of thought and the evolution of language. There are many chapters in the book, including one on the minority rule and one on the Lindy effect, where each one is worth the price of the entire work and deserves to be reread multiple times. Another example of a gem occurs early on in a discussion of the evolution of moral symmetry and why the silver rule beats the golden rule (I won’t spoil it here – buy the book for that discussion alone).

Taleb is at his worst when he is at his most combative. One example is his dismissal of Piketty’s work on inequality. Taleb rightly argues that a dynamic measure of inequality based on ergodicity is far better than a static one. That’s insightful and elegant. But then Taleb goes on to dismiss all of the evidence as if it was just static and as if Piketty doesn’t understand a concept such as ergodicity. Knowing Piketty and having slogged through his book, I submit that both of these assertions are incorrect. A different model of engagement would have been to propose ergodicity as a better measure and then simply ask for evidence. That might open a door instead of slamming one shut.

Slamming doors and picking fights is part of Taleb’s style, however, and is consistent with some of the arguments in Skin in the Game and in Antifragile. A fight gains more rapid exposure for an idea, as the pushback from the other side amplifies the original message (and based on book sales and Twitter follower count that strategy clearly works). A fight also risks reputation and that’s a way of putting skin in the game. That’s likely important to Taleb because otherwise he might leave himself open to the criticism that Skin in the Game borders at times on the kind of advice that he rightly criticizes for, well, not having skin in the game.

This is a good moment to point out that we should all seek out writers with whom we disagree at least some of the time. If we only read books by authors where we agree with every one of their tweets, why bother? What are we expecting to learn? Too many times we are letting our emotional reaction to something an author has said or done stand in the way of engaging with their arguments. Taleb certainly provokes a strong reaction at times, but by all means read “Skin in the Game” nonetheless.

Wednesday, May 2, 2018 - 11:30am

Last Uncertainty Wednesday, I introduced the idea of beliefs. Today we will make this idea more precise. We started with an extreme belief, the one that a coin is so biased that we will only observe “heads” (H). More realistically one might belief that a coin is fair, but has some possibility of being slightly biased in either direction (e.g., more likely to observer H or more likely to observe T).

So how do we formalize this? A belief is simply a probability distribution. For instance the one we just described might be modeled as follows where the horizontal axis shows the values of p from 0 to 1 (where p is the probability of observing Heads)


This is a distribution with a mean at p = 0.5 where the probability decreases to 0 on either extreme (meaning at p = 0 and at p = 1). 

The chart is the probability density function (pdf) for the beta distribution with parameters alpha = beta = 2. You might wonder why the graph goes to values above 1, which would seem to suggest probabilities > 1, but the whole area under the curve is exactly 1. Probabilities are derived from the pdf as small slivers around a value of p (the horizontal axis). For instance between p = 0.45 and p = 0.55 if you imagine vertical lines you get an area of approximately 0.1 * 1.5 = 0.15. So with this belief we are saying that the coin has about a 15% chance of being pretty fair.

Another way to show this belief is via the following cumulative distribution function (cdf):


For each value of p along the horizontal axis, the cdf shows the integral of the pdf from 0 to that value of p, where p = 0.5 represents a fair coin. Looking at it this way we can see how much probability we attribute to the coin being less or more biased than a specific value of p.

Contrast this with a belief that the coin is precisely fair, which would have a cumulative distribution function that looks like this instead:

Here the entire probability is concentrated on p = 0.5! We believe with 100% certainty that the coin is exactly fair. We attribute no probability to it being biased in either direction. The bottom horizontal line ends at 0.5 with a blank circle  and the top horizontal line starts at 0.5 with a solid circle ●, indicating that the value of the function jumps from 0 to 1 at 0.5. So why not draw a probability density function? The reason is that technically we would have to show something different, namely a probability mass function

At a later point we will see just how extreme such a belief is, but even just looking at a discontinuous function should provide an inkling of that.

Tuesday, May 1, 2018 - 7:30am

NOTE: As every Monday, I am continuing to post excerpts from my book World After Capital. Today, starts Part Two of the book, which is title “Getting Past Capital” with an overview and a section on population growth dynamics.

Part Two: Getting Past Capital

Digital technology is shifting scarcity from capital to attention. That is one of the central arguments of World After Capital. With the philosophical foundation out of the way now is the time to back this claim up with some numbers.

First I will examine trends in population growth to show that fears of a further population explosion are unfounded. Then I will look at how much productive capital exists in the world relative to the basic needs of humanity. While that section still needs work, it already contains interesting statistics that suggest we have sufficient capital.

Besides capital, the other critical input to production in the Industrial Age is labor. Labor is provided through what I call the job loop: most people earn a living by selling their labor and then using their wages to buy goods and services, which in turn are produced by other job holders. That loop, which currently captures much of our attention, is being disrupted by digital technologies with important implications for how we could allocate our attention in the future.

Finally, I will argue why attention is the crucial scarcity for humanity going forward. Capitalism, with its emphasis on markets, cannot be used to allocate attention due to intrinsic limitations. Prices do not and cannot exist for the most important activities we should be allocating attention to.


In 1798 Thomas Malthus predicted widespread crises of famine and starvation as population growth outstrips humanity’s ability to grow food [31]. Malthus prediction was half right: Global population did explode, with population growth accelerating right at the time of his writing around 1800.

Since then, the human population has grown from about 1B to over 7B people here on planet Earth [32]. As an optimist, the thing to note immediately, though, is that Malthus’s most dire fears about the implications of this population growth have not been realized. There has been no global scale starvation and even the fear that most people would live in abject poverty has not come true. In fact, the opposite has happened recently. Around the world the number of people living in extreme poverty has been declining all the while population growth has been about twice as fast as what Malthus predicted as an upper limit of 1 billion people added in 25 years [33].

What Malthus got wrong was the rate of technological progress. First, Malthus was wrong in being pessimistic about our ability to improve agricultural productivity. Here is just some of the amazing progress in agriculture since his writing. The percentage of the global workforce employed in agriculture has declined from more than 80% to 33% and is falling rapidly (in the US and other advanced economies agriculture represents 2% or less of employment). Globally in the last 50 years alone, the land required to produce the same output of food has declined by a stunning 68% [35].

Second, Malthus could not foresee the scientific breakthroughs that made the industrial revolution possible. That revolution not only powered the agricultural productivity increase but also gave us dramatic advances in the standard of living, including much increased life expectancy, faster transportation, cheaper and better communication and so on.

Malthus being wrong so far isn’t by itself a guarantee that his predictions couldn’t catch up with us. If population growth were to outstrip technological progress this would in fact be the case. We know this because we have seen it happen in India [36] and other places that have experienced population growth in excess of progress resulting in mass starvation.

As it turns out though, population growth itself responds to technological progress. In particular there is a strong relationship between reductions in infant mortality and decreases in birth rates, as well as between increases in living standards and decreases in birth rates.

Max Roser has produced some beautiful charts as part of his amazing project Our World In Data that show these two effects play themselves out in country after country across the world [37].

So despite the extraordinary growth in global population over the last 200 years, simply extrapolating this growth out into the future would be a clear mistake. Instead, there are strong signs that peak population is a much more likely scenario.

Now one can reasonably argue over what that peak number will be. Some will claim that this debate matters a lot because they strongly believe that the world cannot sustain, say, 11B people. But this misses a crucial point. The world cannot sustain 7B people either—i.e. the current population—if we don’t continue to make technological progress. The way we have managed to support 7B people so far has created all sorts of new problems. We cannot choose to stand still on innovation. Instead we need continued technological progress to solve the problems we have created, such as water and air pollution and climate change.

The key takeaway should be one of curvature. All signs suggest that the global population curve is starting to decelerate (negative second derivative) whereas the rate of technical progress is continuing to accelerate (positive second derivative) [38] [39]. That is the basis for being optimistic about progress in relation to population growth.

I have already described previously why digital technology is so disruptive. Later in the book we will see in more detail how it is contributing to an acceleration of knowledge creation and thus progress. My view here stands in contrast with much of the recent pessimistic writing, including the recently published book by economist Robert J. Gordon and the secular stagnation literature more generally. To show why my outlook is so different, I will now turn to how much capital there is relative to humanity’s basic needs.

Saturday, April 28, 2018 - 7:30am

I have been openly critical of Trump as a president going back to before the election. While I want radical change I do not believe the price for this ought to be going backwards on foundational issues such as the rule of law, press freedom and science. Nonetheless it has been fascinating to observe how Trump’s potential or actual irrationality has opened the door for progress on some issues that were previously deemed intractable, such as North Korea.

It is well known that even in relative simple games, such as repeated prisoner’s dilemma the set of sustainable equilibria grows significantly when there is some possibility of at least one of the actors being irrational (some of the time). In this regard, Trump is a stark contrast to his predecessors such as Obama, Clinton and the Bushes who cultivated an image of themselves as rational actors. For an opponent such as Kim Jong Un, Trump’s (potential/actual) irrationality ironically makes cooperation possible when it was not possible before.

Conversely, continuing to pursue a strictly rational strategy in the face of an actor such as Trump can have disastrous results. That certainly played itself out in the Republican primaries and to some extent in the general election as well. The Democratic party leadership is continuing to operate by “rational” rules, which leads them to meddle in primary elections with the goal of fielding “electable” candidates. This may succeed in congressional elections in the near term but is likely a mistake with regard to their longer term national prospects.

If you want a great science fiction read in which potential irrationality and the equilibria it can sustain is a major plot driver, I highly recommend the Three Body Problem Trilogy.

Thursday, April 26, 2018 - 7:30am

The last few weeks in Uncertainty Wednesday we took a detailed look at the various problems with p-values. Now we will start to work towards an alternative approach based on Bayesian inference. Along the way though we have a chance to think about more foundational ideas, including the subjective versus objective and beliefs versus truths. Today we will start by looking at beliefs and how observations may influence those.

As before with p-values we will use coin flips as they are easy to follow (at least as long as cash hasn’t disappeared entirely). But as we do so, we should keep in mind that this simplicity is actually somewhat deceptive. There is an underlying system that produces the coin flip and the result of heads H or tails T are the resulting signals that we can observe. So when we say something like “I believe this is a fair coin” what we are really saying “I believe that this is a system that will produce an observation of H or T with equal probability.” In fact we are saying even more than that. We are also saying that the history of the system does not matter. No matter what signals we have observed so far, the belief of a “fair coin” means that the next flip could be heads or it could be tails and each with probability 0.5.

Now to start developing an intuition around beliefs and their updates let’s start with an extreme: you belief with certainty that only H will be observed. For instance you might belief the system is “stuck” or in the case of a human flipping a coin you might belief that they chose a coin that simply has heads on both sides! Now you start observing. When you see H, what do you learn? Very little (note: we will eventually see that you have learned nothing). After all, you are observing what you already believe you should observe.

But now imagine that you see tails instead! This should result in a change in belief. You now should belief that T can occur with some probability. Possibly still small, but no longer impossible. Let’s say for example that you observe roughly 1 T for every 99 H. You might be revising your belief to assuming that T is observed with probability 0.01 and H with probability 0.99. You have definitely learned something.

Let’s look at how this is related to the concept of Black Swans. Suppose you believe with certainty that all swans are white. What do you learn from seeing a white swan? Again very little (technically nothing, as we will see). You already believe that all swans are white. But you definitely learn something when you encounter a swan that is not white. Such as when people observed that in Australia there are swans that are black.

This is the same as the idea that the “absence of evidence is not evidence of absence” (a quote often used by Nassim Taleb). Not having observed black swans yet (absence of evidence) does not imply black swans do not exist (evidence of absence). Not having seen tails, does not mean tails can never occur.

To dig deeper into this we will formalize belief mathematically. We will start to do that next Wednesday. In the meantime though you can ask yourself already based on the above: what is a problem with holding extreme beliefs, such as certainty that you will observe H or that all swans are white?

Monday, April 23, 2018 - 5:05pm

NOTE: This is the last post for the foundations section from my book World After Capital. In it I propose a catalog of basic human needs. As set out earlier much of today’s confusion around scarcity results from our willful conflagration of needs and wants.


The definition of scarcity that I just introduced is based on the notion of needs. To argue that there is a shift in scarcity from capital to attention thus requires an agreed upon set of needs to show that we indeed have sufficient capital. Can we make progress in defining what constitutes a set of basic human needs?

I am not proposing that this is a simple task. What follows should be considered a way of starting a dialogue. A list of basic needs is a piece of knowledge. As such it can be improved over time through the process of critical inquiry. You can critique my list by pointing out flaws, you can also propose changes to my list, or you can publish your own list altogether.

One of the benefits of my approach to writing World After Capital out in the open, and with revisions tracked, is that you can see how my thinking on needs has evolved over time. In an earlier version I tried to group needs into categories such as biological, physical, and social. But the boundaries between those seemed rather arbitrary upon further examination. So in the current version I am distinguishing only between individual and collective needs, where the former will apply to a single human wherever and the latter are the needs of humanity as a whole.

Another challenge in putting together a list of needs is that it is all too easy to confuse a need with a strategy for meeting this need. For instance, eating meat is one strategy for addressing the need for calories, but humans can acquire calories from many other sources.

Individual Needs

These are the basic needs of the human body and mind. Without them individual survival and flourishing is impossible. A single individual has these needs even when isolated, such as traveling alone in a spaceship.

The first set of individual needs comes from keeping our bodies powered, these include:

Oxygen. Humans need on average about 550 liters (0.55 cubic meters) of oxygen per day [28]. The exact need of course varies with factors such as the size of our respective body and the degree of physical exertion. Our most common solution to this need is breathing air.

Water. We need to drink on average between 2-3 liters of water per day to stay hydrated [29]. Again various factors such as body size, exertion and temperature will affect the exact need.

Calories. To power our bodies we generally require between 1,500 and 3,000 calories per day, again depending on body size, activity level etc [30]. We solve this need by consuming food. The best way to do this, however, is surprisingly controversial and poorly understood for such a basic need. In particular, the mix between proteins, lipids and carbohydrates is subject to ongoing debate.

Nutrients. The body cannot synthesize all the materials it requires, including certain amino acids, vitamins and minerals. Therefore these must be obtained directly as part of our nutrition. This too is an area that is surprisingly poorly understood, meaning which nutrients exactly we really need to acquire externally seems unsettled. There is a wide range of food consumption strategies that seem to support the human body.

Discharge. So this may be a bit gross but we also need to get things out of our bodies again, including expelling processed food, radiating heat and exhaling carbon dioxide. A lot of human progress has come from better strategies for solving our discharge needs, such as public sanitation. For fans of science fiction, like myself, dealing with the problems of discharge is an interesting limit on our ability to cloak ourselves.

The second set of individual needs relates to the operating environment for humans. From a cosmic perspective, humans have an incredibly narrow operating range, which is provided for, without technological assistance, only in a few places even right here on Earth. Here are some of our basic operating needs:

Temperature. Our bodies can self-regulate their temperature within a limited range. We have a need to control our environment to help our bodies with temperature regulation. Common strategies to meet our temperature needs include shelter and clothing.

Pressure. Anybody who has gone diving knows that our bodies do not handle increased pressure around us very well. The same goes for decreased pressure (one of the reasons air travel is exhausting is that planes do not retain sea level pressure).

Light. Most humans would be hard pressed to do much of anything in complete darkness. The Bible introduces light right away with “Let there be light” in the third verse for the Book of Genesis. For the longest time the solution to our need for light was simply sunlight, but much of human ingenuity has gone into the creation of artificial light sources.

The third set of individual needs arises from dealing with a complex and ever changing environment. As we go through life we encounter challenges that we need to overcome. This results in three fundamental individual needs:

Healing. When we damage our body in some fashion it needs to heal. The human body comes equipped with extensive systems for self-healing including combating many foreign substances (including vomiting, diarrhea, antibodies). Beyond a certain range, the body needs external assistance to heal. Here too we have developed many solutions, and often group them under the term healthcare.

Learning. We are born quite, well, stupid. We even have to learn relatively basic skills such as walking and the use of tools. When we encounter a new situation, we need to learn how to deal with it. We group many of the strategies for solving the need for learning under the heading education, but other solutions include self study, experimenting (gaining experience) and parenting.

Meaning. As humans we have a profound psychological need for meaning in our lives. It is what keeps us going. Religion and religious beliefs have long been a key strategy for solving this need. As I have argued in the section on Humanism, there is an objective basis for human meaning rooted in knowledge. Another key strategy to solve this need comes from our interactions with other humans, including having others acknowledge our contributions to a project or simply our existence.

This last set of needs may strike you as being at a much higher level than the earlier needs. It is tempting to try and sort individual needs into a hierarchy, as Maslov did. That seems intuitively appealing but is misleading. All of these needs are essential. As a thought exercise, picture yourself in a spaceship and try to remove any of the above.

Collective Needs

Our collective needs by contrast arise from living together in societies and sharing space and resources. Meeting these needs is what allows human societies to survive and advance.

Reproduction. Individuals can survive without sex, but reproduction is a need for societies as a whole. As humanity we have already learned how to solve the need for reproduction without sex. In the future there may be altogether different solutions for reproduction in the sense of the continuation of a human society (whether here on Earth or elsewhere).

Allocation. Despite abundance in the digital realm, access to physical objects and resources has to be allocated. Take a chair as an example. Only one person can sit in a chair (comfortably) at a time. When there are multiple people we need a solution for allocating the chair between them. That’s why allocation is a collective need. If you are by yourself you can sit on a chair whenever you want to as there is nobody else to take it up.

Motivation. This may seem like an individual concept but it exists as a collective need in the following sense: Societies need to motivate their members to carry out tasks and follow rules. Even the most primitive societies have solutions for this problem often in the form of rewards and punishments.

Coordination. Whenever there is more than a single human involved in any activity, there is a need for coordination among the participating humans. Take a simple meeting among two people as an example. In order for the meeting to take place they need to show up at the same place at the same time. We have developed many different communication and governance mechanisms to address this need.

Knowledge. As I have argued in the prior sections on Optimism and Humanism, this is the central collective human need. Without increased knowledge a society will encounter problems that it cannot solve and will be decimated as a result. History is full of examples of societies not having enough knowledge, such as the Easter Islanders or the Mayans. This is not about what any one individual has learned but rather about the body of knowledge that is accessible to society as a whole. Much of the later parts of World After Capital are about solutions for generating more knowledge faster.

These collective needs may strike you as overly abstract. But this is the logical result of identifying needs, instead of solutions. Governments and laws, for instance, are examples of solutions to some of these collective needs. But so are markets and firms and more recently networks and platforms.


Now you might ask, what about energy? Don’t we have a need for energy both individually and collectively? It would seem that individually we need energy to maintain the temperature of a house. Or that collectively we need energy to power our communications infrastructure. But as those two examples show, energy is not a direct human need (either individually or collectively). Instead it is an enabler of specific solutions to our needs. Some solutions will require more energy than others.

Here are four foundational enablers. I am listing them in the Needs section, as readers have at times proposed these as additional needs and I had in a prior versions included them among Collective Needs.

Energy. For the longest time humanity relied on direct sunlight as the primary source of energy. Since then we have developed many ways of generating energy, including better ways of capturing sunlight. Producing more energy and having it available in concentrated and highly regulated form via electricity has made many new solutions for human needs possible.

Resources. In early human history all resources were simply found in nature. Later we started both growing and extracting resources. Many modern solutions have been made possible by access to new kinds of resources. For instance, mobile phones give us new solutions to individual and collective needs. Building mobile phones is enabled in part by some esoteric raw materials, such as so-called rare-earth elements.

Transformation. Energy and resources alone are not enough though. To enable most solutions we need to figure out how to use energy to transform resources. This involves chemical and physical processes. Capital, as in physical capital such as machines, has been a crucial enabler for many new solutions to human needs. For instance, a knitting machine can transform yarns into clothing at high velocity. Clothing is one of our key solutions for maintaining the human operating environment.

Transportation. The final foundational enabler is the ability to move stuff (using stuff broadly to include people). This is another area in which we have made great progress over time, going from human powered transportation to animal powered to machine powered, including planes, trains and automobiles.

Again I have chosen these enablers at a high degree of abstraction on purpose. Coal-fired power plants provide energy (in the form of electricity) and so do solar panels today and nuclear fusion at some point in the future. These three examples have dramatically different characteristics but they all are fundamentally energy enablers.

This is my current working version of needs (and enablers). I have now revised this section fairly substantially for a second time. And while I fully expect further changes, I believe it now properly sets up my argument that there is sufficient productive capital in the world for meeting our individual and collective needs, including further development of the four enablers.

Saturday, April 21, 2018 - 7:30am

I recently wrote a post about requiring APIs for social and other applications that have more then 1 million users. Now that is an approach to the problem of market power that adds new regulations. There is an alternative that would go the route of removing existing laws which I want to discuss today.

What prevents someone today from creating their own programmatic control over services? Why can’t I simply write code that interfaces on my behalf with say Facebook? After all, Facebook’s own app uses an API to talk to Facebook. Well in order to do so I would have to “hack” the existing Facebook app in order to figure out what the API calls are and also how to authenticate myself to those calls. Unfortunately, there are laws on the books that make those necessary steps illegal. 

The first is the anti-circumvention provision of the DMCA. The second is the Computer Fraud and Abuse Act (CFAA). The third is the legal construction that by clicking “I accept” on a EULA (End User License Agreement) or a set of Terms of Service I am actually legally bound. The last one is I believe a civil matter, but as far as I know criminal convictions under the first two carry mandatory prison sentences.

So if we were willing to remove all three of these legal obstacles, then hacking an app to give you programmatic access to systems would be possible. Now people might object to that saying those provisions were created in the first place to solve important problems. That’s not entirely clear though. The anti circumvention provision of the DMCA was created specifically to allow the creation of DRM systems for copyright enforcement. So what you think of this depends on what you believe about the extent of copyright.

The CFAA too could be tightened up substantially I believe without limiting its potential for prosecuting real fraud. The same goes for what kind of restriction on usage a company should be able to impose via a EULA or a TOS. In each case if I only take actions that are also available inside the company’s app but just happen to take these actions programmatically (as opposed to manually) why should that constitute a violation?

Sadly I am not optimistic that we will get these kind of changes or anything close to it. We are generally terrible at removing or even just tightening laws once we have them. So even though I would love to see us do this and write about it in the Informational Freedom section of my book World After Capital, it seems more likely that we could get new regulation requiring API access. 

Wednesday, April 18, 2018 - 11:40am

Today’s Uncertainty Wednesday will be the concluding post in my mini-series on the problem with p-values. We have already seen that it is much easier than expected to reject a null hypothesis if you have incentives to do so. We also saw that the ability to work backwards and generate hypotheses from the data is a big issue. Today we will consider a more foundational, epistemological problem with p-values: what is it that we are really learning when we are rejecting a null hypothesis?

Let’s once again consider the original example of a coin toss where our null hypothesis is that the coin is fair (and independent). We have done everything by the book. We had our null hypothesis ahead of time (not generated from the data). We did exactly 6 tosses and they all came up as heads (or tails for that matter), instead of cheating on our data collection. And so with great satisfaction we reject the null hypothesis at a p-value of 0.03125.

But what does that actually mean? What have we really learned from doing so? Our null hypothesis here is incredibly narrow. It is that the coin is precisely fair. Rejecting that leaves open a ton of other possibilities. Is the coin just slightly unfair or is it extremely unfair? Which of these two possibilities is more likely given what we have observed? And why did we pick this narrow null hypothesis in the first place?

Let’s take a step back. Suppose I don’t tell you that we are dealing with a coin, just with a process that has two possible observable signals H and T. If you know nothing else about the process, that allows for anything from observing only Hs to only Ts to some random mix of the two. That makes it clear that having as your null hypothesis that the mix will be random at exactly 50% Hs and 50% Ts is an incredibly narrow assumption. It is picking a single real number, 0.5, on a continuous interval from 0 (no Hs) to 1 (all Hs).

This is related to the issue we encountered previously with spurious correlation. A null hypothesis of zero correlation between two variables is an incredibly narrow assumption, when possible correlation is a continuous interval from -1 to +1. So again, when we reject that narrow hypothesis what have we actually learned? Only that some very narrowly defined assumption is unlikely. That’s not a lot of learning.

This is a fundamental limitation of the p-values approach. Generally people tend to pick very narrow null hypotheses and rejecting them doesn’t tell us much about the alternatives. Now this can be seen as a slightly unfair criticism. If you get a p-value of 0.0000001 on a coin toss and you do it with a large number of tosses you have the information that the coin is likely to be very unfair. But with the p-values approach that additional step tends to be buried.

What is the alternative? The alternative is to take a Baeysian approach instead. We saw that already in the case of correlation how that provides a lot more information than the rejection of a null hypothesis.

Monday, April 16, 2018 - 11:30am

NOTE: I am continuing to publish revised sections from my book World After Capital. Today’s section provides a technological definition of scarcity (instead of an economic one) and provides a brief history of how scarcity has shifted over time from food to land to capital and is now shifting to attention.


In this book I will be arguing that capital is no longer scarce but that attention now is. Furthermore this constitutes the third major shift in scarcity in the history of humanity. The first shift was from food to land when we went from the Forager Age to the Agrarian Age. The second was from land to capital when we went from the Agrarian Age to the Industrial Age.

The words scarce and scarcity have come to take on a meaning that is derived from modern economics. Many people now think of something as scarce if its price is greater than zero. By this definition land is obviously still scarce as it costs a lot of money to buy a piece of land. And financial capital is still scarce because even in our current low interest rate environment, there is a price for borrowing money or raising equity financing (which makes it possible for me to make money from being a venture capital investor).

There is a fundamental problem with this price based definition of scarcity though: anything can be made scarce by assigning property rights. Imagine for a moment that ownership of the world’s atmosphere belonged to Global Air Ltd (GAL). Now GAL could charge anyone who breathes air a usage fee. Air would suddenly be scarce. That may seem like an extreme example at first. Yet, some have argued that the solution to the problem of air pollution is to assign ownership rights to the atmosphere, on the theory that this will result in the owners having an economic incentive to maintain an unpolluted atmosphere.

I will use a different meaning of scarcity that is not based on price. Something is scarce when there is less of it than we need to meet our basic needs. If people are starving then food is scarce.

One can think of this as technological scarcity (as opposed to economic scarcity). The point is that technological progress makes things less scarce over time. The 18th century scholar Thomas Malthus was not wrong about global population growth, which he predicted could be exponential (and thus, he argued, would outpace growth in the food supply leading to hunger) [24]. He turned out to be wrong about the potential for technological progress to exponentially increase the amount of food we could produce. We have in fact gotten so good at agriculture that the amount of land needed for food production has started to decline even as the global population is still growing.

But what about wants? If people are not starving but want more food doesn’t that mean food is still scarce? Is it possible to make a distinction between needs and wants? Modern economics has thoroughly equated the two, but intuitively we know that this is not the case. You need to drink water, but you want to drink champagne. You need to provide your body with calories, but you want to eat caviar. There is no bright line as the use of “starvation” above might suggest—we know that some food is healthier for the human body than other (although we are a surprisingly long way from understanding nutrition well). Still, the distinction is clear enough for this definition of scarcity to make sense. One may argue about degrees but not about the principle.

Just because something is no longer scarce doesn’t mean that it is abundant. Instead there is an intermediate stage which I will call sufficient. For instance, there is sufficient land to meet everyone’s needs for housing and food. For something to be abundant there has to be enough for everyone’s needs to be met at zero marginal cost. Building housing and growing food still incurs significant marginal cost and hence these are not abundant. I am saying “still” because technological progress could make land and food abundant (imagine how much land we’ll have if we can figure out how to live in space and make other planets habitable).

Is anything abundant? Yes, digital information is already abundant. We can make copies of it and distribute it at zero marginal cost. We can meet everyone’s information needs at zero marginal cost.

Is anything scarce? Well, I will endeavor to show that human attention is scarce. It turns out to be scarce, in part, because digital information is abundant.

A Brief History of Scarcity

Food was the original scarcity for humans. We started out as hunter gatherers (foragers). And bad hunters at that. Before the development of weapons and tactics we were mostly hunting small animals and scavenging otherwise. There was one relatively simple solution to food scarcity: migrate elsewhere. And that’s why humanity spread across the globe at a relatively decent speed. But once the human population grew past a certain density and migration was not an option, then food scarcity was the source of much violence both among and within tribes. It is important to note that tribes that were not in direct competition with others for food and had no systems for food surplus (no storage, so called “immediate return” societies) tended not to be violent [25].

Eventually, as far back as 10,000 BCE, we happened upon a series of technological advances including growing crops, irrigation and domesticating animals, that together gave us agriculture [26]. With agriculture, scarcity shifted from food to land (of course land had been a proxy for food to some degree but now the scarcity was land directly). Agriculture increased the food density of land by at least an order of magnitude [NEED CITATION]. That was enough for a meaningful surplus to be produced, which meant that a social hierarchy could be created. Rulers commanded armies. The more land a ruler controlled the bigger an army the ruler could afford, which brought us several thousand years of empire building among agricultural societies. The transition into the Agricultural Age was extremely violent with most forager societies wiped out altogether.

Then sometime in the 18th century a new set of technological advances began to emerge that together gave us industry, including steam/electrical power, chemistry, and mechanical machines. With these, scarcity shifted from land to capital. Why was land no longer scarce? Because the use of machines in harvesting and the increasing knowledge of fertilizers dramatically increased crop yields. The transition from the Agricultural Age into the Industrial Age wound up being incredibly violent with numerous revolutions and culminating in World War I and II.

At the end of the Agrarian Age, the ruling elites all came from controlling land. They still believed land to be the critical scarcity and saw industry as a means of building and equipping more powerful armies. For them industry did not mean a new age had started, instead it meant tanks and battleships. Even World War II was still about land, as Hitler and the Nazis pursued “Lebensraum” (literally: room to live). Once again the transition from one age to the next was brought about through extreme violence. It was only at the end of World War II that we truly exited the Agrarian Age.

We now live in the Industrial Age. Eventually we added service jobs to manufacturing but that did not shift the dominant scarcity which was capital. The success of the market based economy over the planned economy is the result of more effective capital formation. Competitive markets combined with entrepreneurial activity were better at allocating and accumulating capital.

Capital these days is frequently mistaken for wealth or financial capital, but what really matters is productive capital in the form of machines, inventories of goods, buildings. Financial capital is an intermediary step that allows for the formation of physical capital but it does not add to the production of goods and services directly (machines are not made of dollar bills). Companies only require financial capital because of their working capital needs, which arise when they have to pay for machines, supplies and labor before they receive payment for their product or service.

Much like the ruling elites at the end of the Agrarian Age came from land, the ruling elites today come from capital. They often don’t take up political roles themselves, as we have devised ways of influencing policy indirectly, which exposes the owners of capital to less personal risk. A good example of this recently is the role of the Mercer Family in financing and supporting groups, such as Breitbart news, that influenced the outcome of the U.S. Presidential election [27].

The first major claim of this book is that capital is no longer scarce (in the technological sense defined above). We have sufficient productive capital to meet our needs for housing, clothing, transportation, education and healthcare. This is not a claim that productive capital or access to it are adequately distributed around the world. It is also not a claim that we cannot substantially further improve productive capital by making more of it and creating better versions. It is not even a claim that financial capital is currently being allocated properly for the creation of global productive capital (it is not). It is simply the claim that productive capital is sufficient for meeting humanity’s basic needs.

At the same time, digital technology has massively expanded the space of the possible. Digital technology gives us a global network connecting all of humanity to each other and to information at zero marginal cost. Powerful general purpose computing is making artificial intelligence a reality for the first time. This combination of zero marginal cost and universality of computation can dramatically accelerate the creation of knowledge in the world.

Human attention, however, is fundamentally limited. We have 24 hours in the day. We need some of that time to eat and sleep. So that puts a hard limit on how much attention we have both individually and collectively (with population growth slowing down as a result of economic progress).

But why does that make attention scarce? How do we not have enough attention to meet our needs? This is the second major claim of the book. Individually, it is so because most of us are not spending nearly enough of our attention on the fundamental question of our purpose in life. Collectively, it is so because we are not spending enough of our attention on species level risks, such as climate change, asteroid strikes, infectious diseases and opportunities such as space travel, quantum computing, genetic engineering. We are also not paying nearly enough attention to democracy, to our communities, and to each other, including our friends and families.

Therefore the goals of this book are to convince readers, first, that scarcity is, in fact, shifting from capital to attention and, second, that we need new regulation and self-regulation in response to this shift.

Ideally, World After Capital contributes to a dialog that helps avoid another terrible transition. To enter the Knowledge Age we need a lot of changes that are not in the direct interest of the owners of capital who largely control policies at the end of the Industrial Age. This is a direct parallel to the end of the Agrarian Age, and we must learn from that transition, if we do not want to repeat its horrors.

Historians will have a lot of bones to pick with the preceding highly abstracted account. The periods didn’t unfold as neatly and there were regional differences. Nonetheless, I think the overall pattern of scarcity shifting from food to land, from land to capital, and finally from capital to attention holds.

Friday, April 13, 2018 - 11:30am

The Zuckerberg hearings were predictably pretty weak. First, the format does not lend itself to a deeper inquiry on a complex topic. Whenever someone seemed close to getting somewhere, the next questioner would take it somewhere else (or weirdly start over). Second, the politicians asking the questions appeared for the most part not to have a super deep understanding themselves which means they stick to the questions prepared by their staff even when an answer provides a small opening that they should pursue more deeply. Third, the most evasive answers seem to be made possible by the distinction between Facebooks the service and Facebook the company. For instance, when Zuckerberg says that Facebook does not track user behavior on other applications that is true for Facebook the service, but not true for Facebook the company (which owns the Onavo VPN service). Fourth, and in my view most important, privacy is not the critical issue but rather market power which was brought up only a few times.

My fear is that we will wind up with some complex piece of privacy regulation that doesn’t do anything about actually shifting power back from Facebook (and for that matter other larger services such as Google and Amazon) to endusers. That is we are headed for the classic regulatory mistake of regulating the behavior of companies directly instead of changing the market structure so that many possible behaviors can be explored. I have written about a simple regulation that would accomplish a dramatic shift: require services with more than 1 million users to have a full features enduser API. I am quite sure that Facebook would nearly instantly offer a paid subscription alternative to the free advertising model, and even in the free version would be much more mindful of maintaining enduser trust (because now endusers have real power). Instead, the kind of privacy regulation we are likely to get will further cement the existing strength of the large players (who already have a ton of data) and make it effectively harder (rather than easier) for competition to emerge.

Wednesday, April 11, 2018 - 11:30am

Today’s Uncertainty Wednesday continues our exploration into p-values and why they are problematic. Last Wednesday we saw that if you have incentives to reject a null hypothesis, it takes less work than you would initially think to find data that gets you there. I ended that post suggesting that the problem is even bigger than that. How so?

We now live in the age of “big data” – researchers in many fields have access to massive data sets. This lends itself to an approach that has become known as “data dredging.” Instead of starting with the null hypothesis of a “fair and independent coin” we start with a large database of pre-recorded coin flips. Now we work backwards to find a hypothesis that we can reject with a p-value of 0.05 or maybe even 0.01 in our data set!

How would we do such a thing and what would such a hypothesis look like? Well with a dataset containing just Hs and Ts we would have to be a bit creative. But we could generate hypotheses that take the form of a probabilistic finite state machine. For instance: the coin first has a probability of 20% H and 80% T, if H it has a subsequent probability of 70% H again, but if T then it only has a 10% of repeating T. You get the idea. You could write computer code that generates such hypotheses until you find one that you can reject with a really significant p-value in your dataset. Then you go and publish!

Now you might object: Albert, these are completely arbitrary hypotheses, why would anyone believe these? Well, they only come across as arbitrary because I on purpose stayed within the domain of a coin flip. But most big dataset are really complex containing many different variables. Just take the coin flip database and combine it with a database of stock price fluctuations. Now you can test tons of different hypotheses of the form: price movements for stock x are not correlated with the coin flips (where H might be stock price for x moves up and T it moves down).

Again you can have your computer generate these hypotheses for you and test them until you find one you can reject with a p-value that’s deemed significant. These hypotheses are just as arbitrary as the coin state machines I suggested above, but they don’t look that way. They look really simple and thus credible.

But this approach completely violates the statistical reasoning behind p-values. That reasoning only applies if you start with the hypothesis and then apply the test. In any large dataset you will always be able to work backwards towards hypotheses that can be rejected *in that dataset*. Just recall the prior posts about spurious correlation.

OK, so that’s pretty bad given that so many people have incentives to find hypotheses they can reject so that they can publish a paper or claim that a product is effective.  But next Wednesday we will look into an even more profound problem with p-values.

Monday, April 9, 2018 - 11:35am

NOTE: I am continuing to publish revised sections from my book World After Capital. Today’s section contains a central idea of the book, which is that the existence of knowledge provides an objective basis for humanism and establishes critical inquiry as its central value.


What then are the values that I am basing all of this on? Where do those come from?

In his book Sapiens, historian Yuval Harari claims that all value systems are simply narratives that are equally valid. He specifically denies the existence of an objective basis for humanism that would support a privileged position for humanity as a species [22]. I will try to convince you that this is not so. If the power of knowledge is the source of optimism, then its existence alone provides the basis for humanism.

Knowledge, as I use the term in this book, is the externalized — recorded in a medium — information that allows humans to share insights and art with each other.

We are the only species on Earth that generates this kind of knowledge and it can be shared over space and time. For instance, I can read a book today that was written by someone else, a long time ago and in a completely different part of the world. This does give humanity a privileged position among the species because knowledge turns out to be extraordinarily powerful. And to quote from a great tract of philosophy, “Spiderman,” with great power comes great responsibility (which gets its own section later in the book). Because we have knowledge, humans are responsible for dolphins, not the other way round.

Since the work of Alan Turing we know that there is a mathematically precise way in which knowledge gives humans this privileged position. Human brains are more complex than animal brains but they are still only finite state machines, admittedly with a huge number of states. The computational capabilities of finite state machines are quite narrow. For instance, one cannot build a finite state machine that recognizes palindromes of arbitrary length [23]. To get a feel for the limitations of the brain by itself, think about the times you simply cannot remember something and wind up looking it up online.

In addition to our brains though, humans also have universal alphabets and the technology for recording and disseminating information encoded in those alphabets (universal in the sense that once you have an alphabet with at least two letters you can in principle write down anything). This gives humans the same computational capability as the so-called Turing machine which I introduced earlier in the Universality section of the Digital Technology chapter. As Turing showed, that means humanity can compute anything that can be computed in the universe. The computational capability of other species is dramatically limited by comparison. Because they do not have knowledge they are constrained to the equivalent of finite state machines.

Now even if you do not buy into this argument based on a mathematical proof, consider the ability to make progress as a species. Without knowledge (as defined above) other species are reduced to only two methods of sharing something they have learned: communication and evolution. Communication is limited because it is both local and ephemeral and evolution is extremely slow. In contrast, humans can share knowledge across space and time and can rapidly refine knowledge through the process of critical inquiry. What evolution is to DNA, critical inquiry is to knowledge: a process of mutation and selection that over time separates good ideas and good art from bad ones.

Progress and knowledge are inherently tied together through critical inquiry. We make progress only if we are capable of (over time) identifying some ideas as better than others. Some art as more important. Critical inquiry is by no means linear, as new ideas and new art are not always better. Sometimes we go off in wrong directions in science or fads in art. But given enough time, a sorting takes place. For instance, we no longer believe in the geocentric view of our solar system. And only a small fraction of the art that has ever been created is still considered important today. While this process may take decades (and sometimes hundreds of years), critical inquiry is blindingly fast compared to evolution.

My use of words such as “better” implies the existence of values. But where do those come from? They all flow from one central value of a humanism based on knowledge and that is critical inquiry itself. We must at all times guard the freedom to point out flaws in existing knowledge and to propose alternatives. Imagine how limited our available music would be today if we had banned new compositions after Beethoven.

We should therefore seek regulation and self-regulation that supports critical inquiry. In business for instance, critical inquiry often takes the form of competition in the market, which is why regulations that support the functioning of competitive markets are so important. Both the sections on Economic Freedom and on Informational Freedom will introduce examples of regulation that are aimed at increasing competition in the age of digital technology. Individually, critical inquiry requires our ability to be open to feedback in the face of our deeply rooted confirmation bias. This will be addressed in the section on Psychological Freedom. In politics and government critical inquiry is enabled by democracy which gets its own chapter.

Freedom of speech in this view is not a value in and of itself. It is a crucial enabler of critical inquiry. But we can also see how some limits on free speech — which are part of such regulation — flow from the same value. If you can use speech to call for violence against individuals or minority groups then you can use speech to suppress critical inquiry.

Digital technologies, which include a global information network and general purpose computing which is bringing us machine intelligence, are dramatically accelerating the rate at which humanity can accumulate and share knowledge. But these same technologies allow for individually targeted manipulation and for propaganda at global scale as well as constant distraction.

Put differently, digital technology massively raises the importance of critical inquiry, the central value of knowledge based humanism.

Friday, April 6, 2018 - 5:05pm

Earlier this week Rebecca posted the latest evolution of the USV thesis. It includes the words “trusted brands” which led some people to ask how an early stage company could already be that. I provided the answer in a tweet in which I wrote that

“trusted brands” — have incentives that are aligned with your customers/endusers so you can build trust over the longterm

I want to explain that further today and put it in the larger context of Internet business models.

During the Dotcom Bubble, there was a lot of talk about how web sites were aggregating “eyeballs” with the idea that these would be monetized later. Implied in the word “eyeballs” was that the business model would be advertising. And the idea of importing this business model made a lot of sense: after all what companies were doing on the web was “publishing” and advertising had been a key component of the publishing business model for a long time.

But over time a deep problem emerged with this imported business model. Online there was a glut of publishing and attention became highly fragmented. That meant that the price of advertising declined precipitously which put a premium on reach and on targeting. If you wanted to have an advertising based business you needed a large audience and you had to know as much as possible about them to make highly targeted ads possible. But both of these have proven to put companies into some degree of conflict with their endusers. Exhibit A for this conflict today is obviously Facebook with Twitter not far behind.

Interestingly and less obvious, there is also a conflict with advertisers also. How so? Because platforms can retain their value only if they prevent the advertisers from establishing a lasting direct connection with their customers that exists separate from the platform. It is worth reminding ourselves that in the pre-Internet days there was no scalable way for a company to have an ongoing direct relationship with their customers. If you wanted to get a message out you had to avail yourself of media through advertising. You couldn’t just send an email or a text or have visitors to your own website. Catalogs were expensive to print and distribute. And other than getting calls into a call center almost all communications was one-way. Today companies can have their own website and can have bi-directional communications with their customers. But if they did that successfully they would no longer have nearly as much need for platforms (other than for possibly reaching new customers).

Is there an alternative? The most promising model to have emerged are consumer subscriptions. We are already seeing some real breakouts, especially in media with companies such as Netflix and Spotify. At USV we have a growing portfolio of consumer subscription service companies, including Soundcloud, Splice, Nurx, Clue, Duolingo, Quizlet, Codecademy, Skillshare, Stash and Wattpad (some of these companies have hybrid models where they have some advertising for users who do not pay but also a paid version that is ad free).

The consumer subscription model provides a great alignment of incentives between endusers and the company. The company needs to provide enough value in their service so that customers continue to subscribe. Conversely the customer has an incentive to use the system and incurs no marginal cost in doing so. For instance, if you have a Skillshare subscription watching another video or even taking another entire class has no additional financial cost to you. That’s great because it means you can try out a class and if you like it continue. It doesn’t  matter at all how much value you get out of this specific class. All that matter is that you get enough value across all your Skillshare activity.

There is important research that first predicted that subscriptions would turn out to be powerful and important as far back as 1996. I encourage everyone to read Yannis Bakos and Erik Brynjolfsson’s paper “Bundling Information Goods: Pricing, Profits and Efficiency” (PDF) which makes a terrific case for consumer subscriptions.

So I am excited to see the growth of this new model. In a world of direct two-way communications between companies and customers we all need less advertising and more subscriptions.

Albert Wenger is a partner at Union Square Ventures (USV), a New York-based early stage VC firm focused on investing in disruptive networks. USV portfolio companies include: Twitter, Tumblr, Foursquare, Etsy, Kickstarter and Shapeways. Before joining USV, Albert was the president of through the company’s sale to Yahoo. He previously founded or co-founded five companies, including a management consulting firm (in Germany), a hosted data analytics company, a technology subsidiary for Telebanc (now E*Tradebank), an early stage investment firm, and most recently (with his wife), DailyLit, a service for reading books by email or RSS.