Growing the World Population Will Not Lead to More Innovation
Elon Musk argues that we need higher population to innovate more. Here is why he is mistaken.
Today, billions of people live hand to mouth. They are barely scraping by, living a life with too little food and too few opportunities. This condition has been reality for much of human existence. Raising billions of the world's population out of poverty is a challenging task. Now imagine somebody coming around stating: "The solution to all these problems is to add another billion poor people who cannot feed themselves. This act will drive innovation through the roof and solve the whole problem."
We ought to laugh at people suggesting something that nutty, but that is in fact what a lot of seemingly smart people suggest today, including tech billionaires such as Elon Musk. He has elaborated on this particular theme with a variety of tweets. Here is one example:
I know what you are thinking: "You are misrepresenting Elon Musk! He never said he wanted a billion more poor people."
Okay, so what does he want then? That is the problem with the more-people-solves-everything crowd. They are never very specific. Let us be fair and assume that what Elon Musk wants are more affluent and well-educated people in the Western world. What he wants is more privileged white people. No, wait, scratch that. I should not put words in his mouth. We don't actually know what Elon Musk meant, but that doesn't really matter.
If what is needed is more well-educated people who can contribute to innovation and human prosperity, then we don't need to grow the world population. We already have billions of poor people who can get "upgraded" to affluent and well-educated middle-class people. If that is what is required, then we don't need to ramp up baby production. Instead, we could either accept more immigrants to the rich world and thus convert poor people into rich people. Alternatively, we need to channel more resources into poor countries to grow their economies and increase opportunities and education is these countries.
Let me hammer this home: The world is not facing a shortage of people! We have an abundance of them. What we lack is opportunities for all these people. That should be the focus, rather than ramping up baby production, as Elon Musk seems to believe is required to secure the future.
When discussing the impact of population growth on innovation, I have noticed people have a long list of misconceptions about how innovation works, which I want to tackle in this article. Let me list a few key points before delving into the details:
Societies have limited resources to educate people. There is an optimal number of students based on the number of teachers and school buildings available in any given country.
The law of diminishing returns, which suggests that as the number of people grow, we will be less able to employ them effectively in solving the challenges of society. Likewise, as resource consumption grows, we will have to increasingly resort to land of marginal productivity.
Most problems are not trivial to parallelize, meaning you cannot linearly reduce time to solve a problem by throwing more people and money at said problem.
The Challenge of Improving Human Capital
Ultimately, our prosperity is based on raising human capital. It is our investment in people at every level which has helped increase scientific and economic output. It is not merely about the school system but also improved childcare, better nutrition, better dwelling space and so on.
There are several ways to look at this challenge. I will use some extremely hypothetical examples to highlight the problem from different angles. Ask yourself this question: What poor family is most likely to be able to give their kids a good education – The family with two children or the one with twelve kids living in a two-bedroom apartment?
We could take another example: Imagine the number of school children doubled tomorrow. How do you think that would impact education and the well-being of all those children? Consider this fact: Adding more children doesn't make school buildings and teachers magically spring into existence. Such important infrastructure is provided by the adult population. Naturally, society would allocate more resources for schools, but that would only be possible by reducing efforts in other areas of the economy. It is also highly unlikely that we would be able to give every child the same care and attention as before a doubling of the school population was done. That applies even if we knew this doubling would happen many years in advance and could train more teachers and build more schools.
It is not merely a question of limits of physical resources such as school buildings and roads, but also limit on human resources. There is no bottomless well of high-quality teachers in a given society. Whether you seek to have more teachers or produce more oil, you hit upon the concept of the law of diminishing returns. The idea that quality falls at the margins.
The Law of Diminishing Returns
All resources we use come in different qualities. Some agricultural land has higher soil quality and better climate, giving higher yields. Some mines have higher ore grade than others, giving more metal for less energy and money. The same applies to people. Some people are smarter or more talented than others. Within Economics, this fact is central in numerous types of analysis. For instance, market price is believed to be related to what we call marginal cost.
Companies don't sell products at cost of production, but rather at a price which would make them break even if they increased production by one unit. Say the first 9 items made costs $10 each to make while the last 10th item costs $100 to make then the price of every item sold will be slightly above $100 otherwise it is not worth making the 10th item. Hence, production is increased until you reach a cost which push the price to a level that customers are not willing to pay.
Costs can grow like this for many reasons. Increasing production may require paying employees more overtime, which will increase unit costs. It may mean buying more raw materials. The raw material provider may have the same challenge, with marginal costs leading to higher unit costs for a bigger quantity of raw materials.
For instance, growing food production may require using lower quality farmland, which requires more fertilizer and more work while still getting smaller yields.
A good illustration of this phenomenon can be seen with the historical oil price. You might find it surprising that oil prices were as low as $5 per cubic meter in the 1880s, while the last few years it has been over $400 per cubic meter. How can it be that with the primitive technology of the late 1800s, they were able to extract oil nearly 100 times cheaper than today?
The simple reason is that people then had access to much better oil fields, which have since become depleted. As the richest oil fields have become depleted, we have had to use low-quality oil fields at the margins. My own native Norway is a good example. Drilling for oil in the North Sea is hardly optimal. Weather conditions are harsh, and the oil is very deep down. One must build large, complex oil platforms costing billions to get to the oil. Workers and supplies must be brought in with helicopters. Again, not particularly cheap. Earlier, oil exploration could be done on land and required drilling at very shallow depths, which significantly reduced drilling costs.
However, oil is hardly unique. In only ten years, the ore grade of copper mines have decreased by around 25%, which leads to more effort required to get the same amount of copper out. Energy use in copper mines have increased with 46% despite production only increasing by 30% (Guiomar Calvo, Sep 2016).
This problem will only compound itself as energy requirements don't grow linearly with ore degradation, but close to exponential:
When the ore grade decreases in the mine, the energy required for metal extraction increases. For this reason mines with higher ore grades are exploited first, leaving the remainder for the future, hoping that technological improvements will offset those costs. But even if technology improves, the exponential character of the Second Law that can be observed in the figure clearly shows that when the ore grade approaches crustal abundance, the energy needed is exponentially higher. Thus, technology can improve extraction but cannot reduce the minimum energy required for the mining process as the minerals become dispersed.
Thus, the fundamental problem with growing our population is that the cheap resources get depleted, and we are left with lower quality fields and mines which are far more expensive to extract resources from. You can offset that problem somewhat with innovation, but since energy requirements grows exponential in mining as ore degrades, you simply cannot innovate fast enough. Innovation does not grow exponentially with the number of people. Quite the contrary.
The problem applies to all human resources. With a higher population of school age children, we would need to use more teachers of lower skill, which means worse school outcomes. If you hire more teachers without having significantly higher budgets, then that means lower salaries to teachers which will discourage the more talented individuals from seeking the profession, further worsening the situation.
No More Low-Hanging Fruit
Science and engineering have already picked the low-hanging fruit. Further advancement is significantly harder to achieve today. The Atlantic magazine has an article called Science Is Getting Less Bang for Its Buck, exploring this very topic in detail. They have the following graph showing how funding to Ph.Ds have increased along with publication count.
At a glance, that graph looks great, except it does not reflect the actual impact on scientific discovery. Another graph from the same article shows what physicists themselves have rated as the more important discoveries by year. We are not even seeing anything after the 1980s in this bar chart because very little after the 1980s get rated as significant by scientists.
The effect is visible across the board. It is reflected in how economies grow much slower today than in the past. If you look at productivity increases for different decades, you will notice some considerable differences. From 1950 to 1970 Sweden and the US each increased productivity by respectively 117 and 68 percent.
For the decades that followed, growth slowed significantly. Productivity in Sweden and the US grew by 53 and 36 percent, respectively. If we fast-forward to the last two decades, we see neither country manage to get to even 40 percent.
What allowed so much growth in the past was the spread of technologies such as cars, mass transit, electric power tools, refrigerators, washing machines, airplanes, fertilizer, tractors and other equipment mechanizing farming, mining, and manufacturing. The internet and computer revolution simply has not had the same impact, even if we advanced those technologies profoundly. Netflix, Facebook, and WhatsApp doesn't improve productivity.
What point am I driving at? I am trying to get across that we live in a world of diminishing returns. Innovation isn't going to produce the wonders it did in the past. Profound improvements in technology in the past allowed us to leapfrog declines in ore grades and availability of prime agricultural land. Yet, that ability to counter resource depletion with technology is slipping away from us. Going from hand-based mining to using explosives and massive machines allowed radical improvements in productivity that offset ore grade degradation. That kind of improvement has been exhausted. It isn't possible to improve either mining, manufacturing, or agriculture to the same degree anymore. The low-hanging fruits have already been picked.
Parallelization is Hard
A prevalent but naive idea is that progress can be improved nearly linearly with the number of people tasked with making improvements. Many seem to think that if 1 million people produce 10 inventions per decade, then 2 million people will produce 20 inventions per decade. The underlying assumption is that innovation can be parallelized almost perfectly. That creating something new and better are highly independent tasks which can be done in parallel. An alternative perspective with similar outcome is thinking that innovation is all about a few number of geniuses making the radical breakthrough. With this thinking, we need more people to increase the chance of geniuses popping up and advancing society. It is a variant of the Great man theory:
The great man theory is a 19th-century approach to the study of history according to which history can be largely explained by the impact of great men, or heroes: highly influential and unique individuals who, due to their natural attributes, such as superior intellect, heroic courage, extraordinary leadership abilities or divine inspiration, have a decisive historical effect.
With all respect, this perspective is utter nonsense. If you study the history around almost any innovation or advancement, you will quickly learn that there were always numerous contemporaries with similar ideas or concepts. Usually, it was somewhat arbitrary who become the first to make a certain discovery.
For instance, when Thomas Edison made his light bulb, there were already numerous people doing light bulb experiments. Edison was simply the first to get a practical version that didn't burn out too quickly. Newton is known for inventing Calculus, but Leibniz was developing Calculus at exactly the same time. Alexander Graham Bell and Elisha Gray invented the telephone just about the same year.
Hence, what stands out isn't the unique and isolated breakthroughs by lone geniuses, but all the double efforts. All the people making similar inventions and discoveries in parallel. Thus, a likely outcome of adding more smart people to the mix is that we would simply have gotten more people inventing the same things in parallel. Yes, progress would likely have been more rapid, but not remotely as much as many seem to assume.
But couldn't more people working together on the same problem solve a given problem faster? Of course that helps, but four people don't solve one problem twice as fast as two people. But why is that? Why can we not keep adding more people to do things faster?
The reason becomes clear once you analyze any complex task by breaking it down into multiple subtasks, you will invariable notice that only a few steps can be performed in parallel by multiple people. Consider something as trivial as preparing a meal. One person can set the table while another one is cooking. You can have multiple people cutting up vegetables at the same time. However, adding more people are not going to cook the pasta any quicker. You cannot add ten cooks to cut the pasta cooking time to 1/10th. Likewise, the time it takes to fry meat in the pan cannot be reduced by adding more people. It takes the time it needs.
Furthermore, many steps depend on each other. You cannot fry the vegetables at the same time as you are cutting them. Innovation is no different. New technologies become possible to create because prerequisite technologies have been developed. Once heat engines had been made, we could make cars or trains, but not before that. Once electric motors had been made, power tools became possible. Power tools enabled the assembly line and other innovations.
Let me work through a simple example to clarify the problem. The following is an abstract illustration of a complex task which can be broken into multiple subtasks, some of which are dependent on each other while others are independent of one another. The illustration could represent anything from constructing a car to making a brilliant invention or solving a tricky mathematical problem.
Some subtasks will be possible to do in parallel. In my illustration, you can do subtask A1 to A4 in parallel. That means four people could do the A1 to A4 tasks four times faster than one person could do. Two people could do C1 to C2 twice as fast as one person. However, four people couldn't do the C tasks any faster than two people because the subtasks cannot be broken down further.
Two people could complete all subtasks in five steps, as illustrated below.
What if we double the number of people? Would we cut the time in half? No, we only reduce the time it takes from five steps to four steps.
If we double again two eight people, there is no change in the amount of time it takes to complete all the subtasks. We can see more people speed up completion, but not in a very efficient manner. Project manager Fred Brooks made a similar observation in the 1970s and wrote a book about it called the Mythical Man-Month:
Complex programming projects cannot be perfectly partitioned into discrete tasks that can be worked on without communication between the workers and without establishing a set of complex interrelationships between tasks and the workers performing them.
Therefore, assigning more programmers to a project running behind schedule will make it even later
Technological and scientific progress advance in much the same way as I have illustrated in the abstract. There will be some prerequisite technologies, insights, or discoveries which we could label A1 to A4 required before somebody can make discovery B1. If we had quadrupled the number of scientists in the time of Isaac Newton, it is not obvious that it would have let to much faster scientific progress. An excess number of learned men could not have begun working of lasers or nuclear reactors because numerous prerequisite discoveries would have to be made first. Furthermore, those discoveries could not necessarily be sped up by adding more people.
Anyone within the field of computer science should be familiar with these challenges. Until around the 2000s, we were able to make each individual microprocessor run faster. Since then, we have only been able to increase performance by packaging multiple microprocessors onto one silicon die. This presented significant problems for software developers, such as myself. We had to change how we wrote programs. We had to identify which tasks could run in parallel so that we could increase performance by doing multiple tasks in parallel. This was never a simple problem to solve, and still isn't.
Most computer algorithms are like my illustration. They have parts which can be run in parallel, while other parts depends on other tasks being completed before you can work on them. Different problems are easier to parallelize than others. For instance, it is relatively easy to parallelize the work of a web server since handling requests from different users can be treated as independent tasks.
For computer games, in contrast, it has been hard to split the work up into more than 6 separate tasks. Typically rendering the graphics on screen, checking if objects to collide into each other, simulate physics of movements and AI are tasks which can be run in parallel.
Conclusion
Many tasks are inherently sequential in nature. You can only solve one problem after another problem has been already solved. The idea that we can just throw more money and people at a problem to speed things up falsely assumes that most problems are non-sequential in nature. That they can be solved in arbitrary order and thus trivially parallelized.
Many will object and point out that somebody can work on making a computer game while another person is solving the mysteries of human DNA. That is doing work in parallel. Yes, there are many tasks we can do in parallel. However, if you look at almost any major field in science, you will see that there are already countless people working in parallel on the very similar problems. Nuclear fusion, for instance, would be really cool to have. Despite being rather esoteric, there are numerous research organizations and companies working in parallel on fusion power. That isn't a complete waste because one approach may work better than another.
However, in terms of resource usage, it is still quite wasteful. It would have been far more optimal in terms of resource usage to test one approach after the other in sequence rather than doing it in parallel. That way, you could abort trying numerous approaches as soon as you have found a working solution.
Of course, we are a bit rushing, so we can afford to be a bit wasteful. Modern microprocessors are quite similar. They waste a huge number of transistors to get relatively modest performance improvements. However, our advances in miniaturization has made transistors so cheap we can afford to be wasteful.
My motivation for writing this article has been to make the case that adding more people to the planet will not lead to more innovation, but instead amplify pressure on our limited resources. Thus, adding more people may well lead to the opposite outcome. The low-hanging fruits have been picked, which means resource shortage will increasingly be felt. It is pretty clear that innovation relies more on having an abundance of resources rather than abundance of people. Well-educated people with advance equipment can achieve a lot more than poorly educated people with few tools and cheap equipment. Quality beats quantity.
We are at the margins of innovation today, which means if we want to really feel more economic prosperity, we need resources which are cheaper to access. That will not be the case if we are forced to extract resources from marginal lands. The simple antidote is lower population. Fewer people need less resources, which means you can limit resource extraction to the mines with the highest ore grade, the land with the most fertile soil and extract wind power from the windiest places.
Eventually, we will reach a technological level which allows for strong growth in the future. For instance, outer space offers many opportunities, but being able to utilize resources in space is very far away. It will simply take time to develop the technology, and you cannot take shortcuts by just throwing more people at the problem. Technology simply needs time to mature.