Elon Musk published a few hours this tweet:
“Population collapse due to low birth rates is a much bigger risk to civilization than global warming”
Robin Hanson, who is something for whom I feel a lot of intellectual respect, liked the tweet.
In my model of the world works, overpopulation is in fact a big problem. In general, the more people you have, the less resources you have to share among those people. A decreasing population would be in fact good news, although maybe not in the short term.
Can you help me understand what are Elon/Robin seeing that I am not?
A couple of extra points for the sake of clarity:
-
I do understand that, in the current system, having an aging population is a problem because many resources go toward people that reach an old age
-
AI might or might not end the world. Let’s assume in this scenario that it does not and we have many more decades ahead
The only hard limit on our resources that we currently know of are the energy output of our sun in the short term and the energy output of reachable stars in the long. We are not constrained on food, space, energy, etc on earth in the short term. We have only artificial political constraints where the stationary bandits ruling the various geopolitical units have subsidized the creation of narratives justifying your constraint. Humans do not consume wealth, they create it.
I think this is both trivially true an irrelevant. The fastest car that we can build is only limited by the speed of light, but we can’t really make cars that fast for a vast number of reasons that I am too lazy to spell out. I am not interested in a theoretical world where if we run out of phosphorus on Earth we just make new atoms using fusion or we make rockets to mine distant asteroids. I’m interested in Earth on the 21st century, where if you interrupt the transit of boats in the Black Sea people in Africa die of hunger. We are, in fact limited in the amount of food we can produce, and although we could increase the amount of the Earth surface devoted to crops, that comes at a high ecological cost. Same thing for many other resources.
We are constrained on food. There is a limited among of land suitable for farming, most of which is in use, a limited amount of fresh water, and we are close to ceilings on optimising agriculture with machinery and fertilisers. The Ukranian war is affecting food security, which means there’s no slack.
This is clearly wrong. We only produce enough food to feed the people who are alive right now, and market (and other) incentives mean it’s not worth it to build in lots of additional redundancy into the system which might come at too high a cost.
One way to think of this is in terms of the long-run price elasticity of supply. How much more wheat do you think the US would produce in ~ 5 years if the government increased agricultural subsidies to double the revenue producers get from the sale of per kg of wheat, for instance? I think a reasonable answer is “more than two times as much”. Given that food is a negligible fraction of US gross domestic product today, the additional benefit from having more people is clearly going to overwhelm the increased relative price of food.
I’m not sure that the production of food is so elastic (it could be, I simply don’t know). A few things make me thing that it is not the case. The crops we use are already very efficient and a big fraction of the soil is already being used for agricultural purposes. I don’t know if you have data to backup that we could easily double the production with incentives.
This paper quotes the long-run price elasticity of wheat growing area at around 0.37, though they acknowledge that there’s significant controversy over these price elasticities and the exact values estimated are highly sensitive to modeling assumptions; from what I can gather the paper is actually on the more pessimistic side of the literature.
That’s just the growing area elasticity; it doesn’t take into account other margins on which you can increase production (such as better tech, more capital investment, productivity gains from having more people etc.), but even at 0.37 a doubling of the price should give you a 30% increase in wheat supply.
The paper also cites elasticities for many different kinds of crops, and the price elasticity of supply for soybeans and wheat is significantly higher than the price elasticity of supply for e.g. rice. This heterogeneity further leads me to believe that there’s another margin we can push on, which is to use crops whose supply is more price elastic as overall crop demand goes up. For example, this paper from Brazil estimates a national long-run soybean price elasticity of supply around 0.8, and in some regions of Brazil the elasticity exceeds 1.
I think I still stand by my prediction that the long-run price elasticity of supply for wheat should be well above 0.37 and slightly above 1, but this would come from pushing on many different margins and not just increasing land use.
Thanks, that was very informative! I update towards “it is still possible to produce more food” than we currently are
Of course its worth it...it saves lives. People in the ancient world built grain stores, because they understood that. We switched to a model based on global trade and just-in-time sourcing, which turned out to be fragile in the face of disease and war.
Globally, famines occur, and the war in Ukraine is affecting food supplies in Europe and other places. It’s not affecting the US much, because the US is a long way away, and a big agricultural producer...IE, it’s a special case and not proof that there is always enough food.
Not to be crude, but due to global inequality in income and wealth, it’s quite possible that it’s not worth it to build redundancy to agricultural production in the US in order to save the lives of people in Niger in the event of a global crisis causing disruptions in the supply of food.
When I say “it’s not worth it due to incentives” I mean that it’s not profitable to do so, in the sense that the expected revenue you would get from it is less than the expected cost, or at least you think this is the case. Note that modern financial instruments such as commodity futures and options provide great flexibility to farmers who try to lock in prices in advance or overproduce to meet potential surges in demand.
I never said “there’s always enough food”. Saying that “there are sometimes famines, so food is a binding constraint on increasing population right now” is a bad argument. It’s like saying “the US sometimes has power cuts, so power supply is a binding constraint on increasing population”—it doesn’t take into account the cost of having a power grid that operates at 100% reliability at all times, which could plausibly not be worth it from a profit and loss perspective.
It’s worthwhile for a society to do things that prevent members of that society from dying unnecessarily. I wasn’t speaking about the US specifically. I am not USian. The only thing I had to say about the US specifically was to say that it is relatively insulated from the problem. I was saying that society X needs it’s own reserves because the alternative, trying to ship resources from whatever they are cheap, is just too fragile and unreliable...it’s been tried, and failed.
I know. I know you are using the “dollar profit” definition of “worth”. Thats not the definition of “worth” that I am using.
You also need to consider the question of how many people die unnecessarily in a very wealthy country because of an unreliable power supply. People were freezing death in Texas a few years ago.
I don’t think it’s worth continuing this discussion, as I have neither the energy nor the mental bandwidth to put into it right now. You’re going far away from the original claim about whether having more people would make each person better or worse off and discussing things which are quite irrelevant to that question.
This is the second time in this thread you bail out of a discussion this way, which I find extremely rude. Didn’t downvote it the first time but I will this time. I would suggest that next time just simply refrain from replying.
I find your reaction mystifying. Are you seriously downvoting me because I’m being truthful about why I don’t want to continue the discussion? You think it would be nicer to just ghost the person you’re talking to?
To some extent, yeah. If someone ghosts me in a discussion like this I think “ah yes, clearly I have bested them and they have no reply to make” and I feel good about myself. If they say I’m not worth talking to I feel bad about myself.
Of course it may not be a good community norm to do what’s nicer.
I note that your second sentence was kind of a parting blow, saying you’re not going to argue and then making one final argument.
Fwiw, this is the tactic I often use in situations like this.
My argument is not about the object-level, it’s just an explanation of why I think it’s not worth continuing the discussion. I feel I should provide some explanation of why I want to cut the discussion off here, and for me what matters is whether we’re making claims that in principle could be falsified or not, and whether those claims actually relate to the question asked in the post.
For instance, when I give power cuts in the US as an analogy for why food insecurity doesn’t necessarily mean we’re bottlenecked by food supply in growing population, and I get a response that
“You also need to consider the question of how many people die unnecessarily in a very wealthy country because of an unreliable power supply. People were freezing death in Texas a few years ago.”
I mentally check out of the discussion because the counterparty is not even trying to understand my point and how it relates to the question raised by OP. I find conversations of this nature exhausting.
Yeah, I agree that tactic is better and I could have used it here. Thanks for pointing this out.
I think the tweet is saying a lot less than it seems like it is.
In particular, the risk of civilization collapse due to climate change is really small. There are all kinds of bad things that might happen due to climate change, and I’m strongly in favor of working hard and spending a lot of resources to fight it, but I think it’s < 5% for even business as usual to lead to civilizational collapse due to climate change.
Suppose Musk thinks that as well—and in fact probably his estimate is lower than mine for collapse. Now he’s only needs to think that there’s a small chance of collapse due to low birth rates for it to be greater than the civilizational risk of climate change.
He also said low birthrate is more dangerous than AI (which I think makes no sense).
He also said: “Doing my best to help the underpopulation crisis. A collapsing birth rate is the biggest danger civilization faces by far.” He also followed up the original tweet here with “(And I do think global warming is a major risk)”.
Connecting his statements literally, you come to the conclusion that collapsing birth rate is a by far bigger than a major risk. I think this means that this answer (that he puts both global warming or population collapse to have a quite small probability to collapse civilization) is false. On the other hand, he’s well known to make ridiculous statements on twitter without much regard for the truth.
It’s true that more people means we each get a smaller share of the natural resources, but more people increases the benefits of innovation and specialization. In particular, the benefits of new technology scale linearly with the population (everyone can use the) but the costs of research do not. Since the world is getter richer over time (even as the population increases), the average human is clearly net positive.
This model is wrong because it doesn’t take into account the enormous gains to productivity you get from various kinds of economies of scale, particularly from the possibility of specialization and trade between different people.
Even if you had all of the technical knowledge we have right now, a town with 1000 people would be much poorer on a per person basis than we are now. There would simply not be enough people working on enough kinds of different jobs to make most of what makes the modern world possible work. For example, how do you build a modern laptop with only 1000 people? Even if all 1000 were exceptionally talented, that’s just not enough scale for you to be able to produce a laptop.
In addition, there are large positive externalities from having additional people, because a lot of research is high fixed cost and small marginal cost. It doesn’t cost ten times more in research effort to make ten rockets compared to making one rocket. So the more people you have who are working on figuring out how stuff works and spreading that information around, the more productive everyone becomes.
You can think of this as a tradeoff: the more people you have the more you’re able to produce per person due to the effects mentioned above, but the less natural resources you have per person to go around. If you view this as a purely static problem then there would presumably be some “optimal level of population” at which income or wealth per person is maximized, so the question is how big this optimal level actually is. I think all available signs right now point to us being far below the level of population that would be optimal: the bottleneck on our productivity and economic growth is not resource constraints but the lack of enough people.
I roughly agree with this model, I disagree with “all available signs right now point to us being far below the population level tat would be optimal”. I don’t know why you think that the bottleneck is “the lack of enough people”.
The best explanation I know of why economic growth stopped accelerating circa the 1960s is “we don’t have enough people”. See e.g. Open Philanthropy’s explosive growth report.
In addition to this, predictions based on resource bottlenecks have not done well in the past. The model that “resource constraints are binding on how many people we can have” has consistently done badly at predicting the future over the last 200 years. Given that the model has worse fit to the past data unless you save it by adding an extra parameter, you should update against it in light of its bad track record.
Ask yourself whether a person thinking like you would have made accurate predictions in 1900 if they did not know what would happen in the future, or even better; write down an explicit stochastic model of growth and see what happens if you fit it to data pre-1900 and extrapolate it to post-1900. Which model is able to make better predictions? I’ve done this exercise before and I can tell you that the models that get the best fit to the data are the ones in which population bottlenecks are serious but resource bottlenecks are not.
If you roughly agree with the model, can you model your uncertainty over the optimal level of population as a lognormal distribution and give me the parameters? Why do you think it’s lower than ~ 10 billion people?
First, I would be careful using the argument “models that did not work in the past won’t work in the future”. A person in 1900 would probably fail to predict the development of the key technologies that allowed the population growth we saw in the XX century (the Haber–Bosch process and the crops improvement), the same way we might fail to see potential technologies that will be developed in this century and would allow even higher growth. But we shouldn’t count on those new technologies to materialize: they might happen, or they might not.
Second, I think it would be a futile exercise to fit a lognormal distribution about my uncertainty: treating the world as a single entity does not really make sense in the world we live in right now: Places like Bangladesh or India would be clearly better off with a (much) smaller population and there might be places where it could be better having more people available. My impression is that, in general, most developing countries (the ones where the population is growing faster now) would be much better off with smaller populations. Then, I wouldn’t really know how to estimate what is the “best” number, and even if I did, I would be just engaging in a pseudo-statistical analysis to give a fake sense of certainty. However, even if I cannot give you a precise value for the parameters of a lognormal, I think that if I see a person who is 1.7m tall and weighs 300 kg, I should be allowed to state: that’s too much. I don’t know what his “correct” weight should be, but it is clearly less.
Last, I do agree (as it has been argued before in this thread) that a perfectly managed planet Earth should be able to sustain a way larger population: but we are not a perfectly managed Earth and we probably won’t be at any point. Given that humans are imperfect agents with very limited coordination ability, I would prefer to live in a world where the population stabilizes to 10 billion people or less than in a world that has 100 times more humans.
My exact argument is that if a particular type of model can’t accurately retrodict the past without overfitting then you should update against it when making predictions about the future. I didn’t make an absolutist claim such as “models that didn’t work in the past won’t work in the future”, just that the fact that they didn’t work in the past is an update against them.
A specific country like Bangladesh would of course be better off per person if it were less populated, but that’s just because they are able to trade with the rest of the world and their current level of population is adapted to those conditions. You can’t compare the optimal level of population when you can interact with the outside world with the optimal level when you can’t; they are quite different.
I think this conversation is not fruitful so I’ll stop it here unless you can come up with specific object-level predictions about e.g. the next 10 years that we would disagree about.
To sustain high tech-driven growth rates, we probably need (pre-real-AI) an increasing population of increasingly specialized and increasingly long-lived researchers+engineers at every intelligence threshold—as we advance, it takes longer to climb up on giants’ shoulders. It’s unclear what the needs are for below-threshold population (not zero, yet). Probably Elon is intentionally not being explicit about the eugenic-adjacent angle of the situation.
That is true in general, but we are nowhere near the resource limit currently, provided we manage them properly. There are existing low- to high-tech measures, like reducing cattle population, improving efficiency of existing HVAC systems, switching to electric, GMO crops, improved renewable and nuclear energy sources, smaller dwelling footprint per person, and many others. This doesn’t include any new tech, like fusion, vat meat, energy beaming from space, moon bases etc.
With an eye on conservation of resources, without giving up quality of life, the planet can probably easily support an order of magnitude more people than there are now.
Most of the world is covered with people. Economy and its productivity cannot be measured today. The measures or measuring stick we use to explain the world today is inadeqate. Retrospective analysis of humanity is nice. But very risky business since nothing today ressembless the past.
To add—undercutting the human demographic pyramid has serious social drawback. Capital grows on influx of human population. Unfortunately the idea that human demographic growth is necessary is probably bias based around human psychology and large capital, that maintains its relevance merely by generating profit. If we neglect these rather stupid ideas, we are left with infinite sea of positive options where less population is always, better. We are not few. We are billions.
I have a terse answer: If you’re not growing, you’re dying.
Somebody has to support the elderly people. The other aspect is the capitalism and consumerism point of view. But all this hopefully will be good for the planet and other species.
we need ideas to have econ growth, we need humans to have ideas, having fewer humans will lead to econ stagnation. This is the summary of the argument https://www.cold-takes.com/what-is-economic-growth/
But econ growth does not necessarily mean better lives on average if there are also more humans to feed and shelter. In the current context, if you want more ideas, you’d have a better ROI by investing in education.
Note that decline is not collapse is not extinction.
https://astralcodexten.substack.com/p/slightly-against-underpopulation
If human lives are good, depopulation should not be pursued. If instead you only value avg QOL, there are many human lives you’d want to prevent. But anyone claiming moral authority to do so should be intensely scrutinized.
Isn’t he just trying to win points from his new republican buddies by disowning earlier interest in climate change and aligning himself more with a cluster of socially conservative views (anti abortion, love of big patriarchal families, fear of being outgrown by the out group etc)?