Hide table of contents

People who have looked into the topics of resource depletion don’t consider it to be a direct civilizational risk. Probably because of this, there hasn't yet been a post summarizing this area. Nevertheless, I thought the topic deserves a place in CEARCH's longlist of causes, so I put together the fragments I could find as a directory of related ideas.


  • Running out of resources seems very unlikely to pose an x-risk, even without considering AI progress.
  • If politicians continue to act on climate change, it seems unlikely to increase resource scarcity.
  • Biogas plants and farming technology seem like potential opportunities for impact entrepreneurship.
  • I agree with existing EA assessments of ecosystem collapse that it seems unlikely to pose an x-risk, but I find the issue to be more uncertain than they imply.

Potential problems with waste


  • Some materials are expensive to extract from waste, which could be reducing our chances of restoring civilization from a catastrophe


  • Ineffective use of resources
    • E. g. meat is resource ineffective, so decreasing its consumption makes food in the 3rd world cheaper
    • The (lack of) redistribution of unwanted items, food, clothing
  • Waste pollution has significant health effects in LMI countries, where landfills are badly managed (12)
  • Contribution to wild animal suffering


  • A negative ecosystem impact
  • Greenhouse gas emissions

Past analyses

Resource depletion

The most informative summary I found was David Denkenberger's 80,000 Hours interview, which covers the most discussed scarce resources: Nitrogen, phosphorus and energy. Firstly, nitrogen and phosphorus are becoming scarce in forms, which are necessary to grow plants (such as nitrates and phosphates). Although using them as fertilizers allows us to feed around a two times larger global population, we are currently dependent on artificially producing them to sustain this population. At the same time, both nitrogen and phosphorus production currently depends on non-renewable sources - natural gas for nitrogen and phosphate mining for phosphorus.

Fortunately, Denkenberger and his colleagues have found out nitrogen can be cheaply and sustainably replaced, as it can be isolated from the air using renewable energy, with an efficiency loss of only 1 %. Finding a renewable source of phosphorus seems more tricky, however, Denkenberger mentions even a 10x increase in its price would hardly affect the price of food. Nevertheless, the phosphorus we do "use up" isn't inevitably lost, just more expensive to extract. And Halstead estimates it would take at least 6,000 years to use up all phosphorus reserves, which means that we have a relatively long time to discover technologies which would make its extraction cheaper.

The concern with energy relates to its increasing scarcity, which partly owes to the transition from fossil fuels. Denkenberger mentions this scarcity could be well managed, if solar and wind power are combined with alternative energy storage methods (compressed air, pumped hydro) that just aren't incentivized by the current markets. Nevertheless, there has been an extensive debate, as to whether these methods will get implemented quickly enough to prevent an economic decline.

The optimistic sentiments were also echoed on the podcast by Luisa Rodriguez who wrote a comprehensive report on existential risks from civilizational collapse. Importantly, Luisa concludes that even catastrophes which would erase 99 % of humanity would be very unlikely to permanently halt humanity’s potential. Therefore, resource depletion seems unlikely to pose an existential risk, as all humanity would need to run off a resource, which is irreplaceable and necessary for life. Nevertheless, it may exacerbate other risks that require global cooperation, as I’ll expand below.

Similarly, in a short mention in The Precipice (p. 117), Toby Ord implies he hasn't found evidence that the scarcity of fossil fuels, phosphorus, topsoil, fresh water, or metals would pose a threat to humanity's long-term potential. Ord has also written an article, in which he argues that given the intrinsic value of human life, it would likely be good if Earth was more populated (particularly if resources were distributed more efficiently via social change or technological progress).

In What We Owe the Future (pp. 147-150), MacAskill mentions that having easily accessible fossil fuels may be crucial to allow a civilization recovery after a collapse, i. e. industrialization (presumably - before reinventing cheap renewables, fission, fusion, or AGI).

Lastly, the UN has recently warned about the depletion of river sand used for concrete, however, there seems to be plenty of alternatives and the UN’s report mentions there are viable ways to recycle construction material. For instance, the EU has funded a research project which claims to have managed to develop artificial sand from fly ash which is “superior and cost-effective alternative to natural sand”.


Another potentially important consideration in this area is the threat of increased resource scarcity due to overpopulation. Population growth in the developing world has been slowing down since the 50’s due to the demographic transition. However, the African population is expected to grow until the year 2100, which is also around the time when temperatures are estimated to peak (IPCC 5; p. 1046). I should note here that people typically only tend to choose to have fewer children once mortality decreases (12). Therefore, saving lives in the developing world likely has no effect or reduces this problem (see GiveWell’s report). Additionally, apart from taking a toll on human lives and wellbeing, malaria burdens cognition and the economy, amplifying resource scarcity.

According to a report from 2023, if politicians continue implementing green reforms at their current pace, we are headed for a maximum warming of 2.5°C by 2100. Models included in IPCC 6 and Halstead’s climate report estimate that a climate change of similar magnitude would cause the global GDP to be ~5 % lower in 2100, than it otherwise would be. Importantly, these models estimate that such a decrease wouldn’t be enough to reverse GDP growth, even in Asia and Africa, GDP per capita is projected to grow 5-9 times under “business as usual” approach to global inequality [1].

Mark Lynas discussed 3 resources that could become more scarce as a result of climate change: Fresh water, food, and land with a habitable climate. This could cause migration of 1 billion people by 2050 and reignite regional conflicts in Africa, as well as populism in wealthier regions. Such a hit to the world’s stability could damage global AI governance and increase the threats of global conflicts, engineered pandemics, or dangerous geoengineering (12).

Fortunately, none of the discussed resources are something of which we can suddenly “run out” - in a way that would create a leverage for EA work. Nevertheless, all of these already are scarce resources and some EA work contributes to changing that. Most importantly, ALLFED’s work on alternative food sources also includes extreme climate change scenarios. For water, there are Dispensers for Safe Water. Lastly, Hannah Ritchie is working on the ineffective use of land, as meat production and a lack of investments in sub-Saharan agriculture is making food in Africa more scarce.

However, as living standards are estimated to rise, the EA theory of change in these cause areas might have to rely on the assumption that higher world GDP will lead to better problem solving. While that’s a legitimate perspective (taken by Progress Studies, for one), it’s also possible we live in an age where the impact of making the economy generally more efficient gets to some extent outweighed by the fact it accelerates the availability of dangerous capabilities. This seems like a minor concern for helping the poorest. However, if our goal is to increase the world’s stability, more direct paths seem like a safer & more effective bet.

Additionally, some EAs study the prospects that our climate predictions might be wrong. For instance, one study estimates a warming of 10°C would cause 5 billion people to die. Additionally, although our present understanding suggests triggering continual feedback loop of warming is impossible (runaway warming), Beard et al. state “the dynamics and tipping points of global ecological and social systems also remain highly uncertain.“

Ecosystem collapse

One area that could be an important part of this uncertain dynamics is ecosystem collapse. The 80,000 Hours review of climate change suggests an existential risk from ecosystem collapse is extremely unlikely. However, I’m afraid the CSER study they link in support doesn’t lay grounds for any firm conclusions.

The study provides the following argument: Yes, the data suggest global biodiversity is in decline. Yes, small-scale experiments have shown local biodiversity decline to reduce plant productivity. However, local biodiversity seems to be mostly staying the same, because when one species dies out, species from other regions take over its former niche. Therefore, e.g. as much as 50 % of species could die out without significant damage to human civilization. The authors consider such loss of biodiversity to be plausible, but state they don’t see how this could lead to human extinction. They admit a 90 % biodiversity loss already could pose an existential risk but claim no one predicts such a loss. Lastly, referencing D. Meier’s What’s So Good About Biodiversity?, the study mentions there’s a “growing skepticism regarding the strength of evidence linking trends in biodiversity loss to an existential risk for humans” but also that “the majority of conservation scientists are likely to flinch at this conclusion”. I’ve read Meier’s chapter on “Biodiversity as (Human) Life Sustainer” but didn’t stumble on any relevant arguments there, as it only attacks very straw-man versions of the concern[2].

Although I’m inclined to agree with the general case, I don’t think these arguments are sufficient for the stated level of confidence. Both good counter arguments and a better case against ecosystem collapse as a cause area can again be found in Halstead’s review (Chapter 5.7). 

It's true that if we only look at existing anthropogenic biodiversity loss (1% of species since 1500), ecosystem collapse looks quite unlikely. However, to assess the problem, we should be focusing on future extinction rates, which are expected to be mainly driven by climate change. A 2019 report by the UN consensus (IPBES) estimates “the fraction of species at risk of extinction due to climate change is 5% at 2°C warming, rising to 16% at 4.3°C warming”. Halstead implicitly assumes substantial problems start after 75 % of species go extinct (6th mass extinction), which would only happen after 18,000 years at the current pace. Additionally, Thomas (2015) estimates that the selective pressures posed by humans accelerate the natural creation of new species 100 - 10,000 times, while we only speed up natural extinction rates 100 - 1,000 times.

However, these estimates are riddled with substantial uncertainty. Firstly, a look at population losses tells a much more pessimistic story. Since 1970, wildlife populations have decreased by 68 % on average (this figure is a little tricky). And this loss is likely even higher for "unsexy" species, such as insects and fungi, where 99 % of species haven't been evaluated for extinction - a 2022 study in Nature estimated that 28 % of well-studied species and 56 % of data-deficient species are “threatened by extinction”. [3] Thirdly, some conservationists are concerned about ecological cascades caused by extinction of keystone species.

Nonetheless, an estimate published in 2022 in Nature by Kaiho suggests that if we continue to reduce our GHG emissions (leading to a 3°C warming by 2100), the average animal species loss will reach 10-15 % between the years 2060-2080 or a maximum of 33 % in the worst case. The author suggests this would be sufficient to avert an ecosystem collapse with catastrophic impacts, that could happen if we continued burning fossil fuels or engaged in a nuclear war. However, "ecosystem collapse" is only operationalized as the extinction of 20-50 % of animal species. Although it's a common meme that e.g. keeping the biodiversity loss under 10 % is necessary for civilization, the aforementioned reviews haven't found evidence such a tipping point exists.

Halstead expresses optimism, highlighting the literature that suggests that the relationship between the habitat area and the number of species is logarithmic, meaning that while habitat loss may manifest in significant population losses, we would have to significantly increase our land use to cause the extinction of these species (the species-area relationship). Additionally, while tropical deforestation continues, it has significantly slowed down since the 1980’s and forest cover has been rising in the temperate forests.

Secondly, Halstead investigates the claim that a loss in biodiversity could lead to resource scarcity because of lower crop yields. He remarks there hasn’t been a marked decrease in crop yields throughout Europe’s history despite its loss of 88 % of forest cover between 1,000 BC and 1850 AD - nor a nonlinear ecological collapse. Similarly, he points out that many of the currently richest regions have the Biodiversity Intactness Index below 30 %.

These correlations are somewhat confounded: They don't actually capture a decrease in biodiversity, but forest cover; and rich regions are largely dependent on imports. Nevertheless, the fact that European crop yields have not been affected by the loss of 88 % of forest land, does provide some evidence against the idea that a decrease below 90 or 30 % of biodiversity intactness triggers tipping points that make agriculture problematic. Particularly so, if we combine the findings regarding the non-linear species-area relationship with the fact that forest cover is mainly decreasing in regions with biodiversity intactness > 90 %.

Additionally, I think there are good reasons to expect that in the event of a 6th mass extinction, humans wouldn't be one of the 75+ % of species that would disappear, as we aren't dependent on one habitat or one source of food, and have the unique ability to transform nature into whatever nutrients we need [4]. Since a mass extinction caused by an ecological cascade would likely take at least decades to unravel, it seems likely that at least 80,000 people would survive (which is the upper bar for the minimum number of people to restore civilization estimated by Rodriguez's report), as decreases of human population would increase biodiversity in a self-regulating fashion. Our ability to adapt may be the reason we're the most populous mammals.

In summary, I think Halstead presents good reasons to be optimistic regarding the threat of an ecosystem collapse in this century, although the evidence in this area seems somewhat uncertain. Hannah Ritchie, who has collected much data on the topic, suggests increasing crop yields (particularly in sub-saharan Africa) could again be an effective way to reduce the problem of species loss, as it allows us to use less forest land to grow crops. Further study may be devoted to the impact of keystone species and second-order effects of climate change on agriculture.

Waste pollution

Rob Wiblin has written a post suggesting landfills don't cause significant pollution in rich countries, where they tend to be well managed. He also suggests burning waste can be good for the environment in rich countries, as it also generates energy. As he notes, this certainly isn’t the case in poorer countries, where open trash burning is common (e.g. in Delhi, it accounts to 10-30 % of air pollution). However, all sources I could find suggest recycling is still more energy efficient than burning trash & using it as an energy source (123). The worst claims of the articles with an anti-recycling overtone were “it isn’t as great as it seems” (12).

Waste makes up 3 % of global greenhouse gas emissions and the overwhelming source of waste emissions seems to be methane from organic matter (1234). Good news is that these kinds of emissions can be eliminated and used to create electricity with an efficiency ranging from 25-100 %. In areas where the efficiency is > 50 %, the global warming potential is net negative, because of the energy generated. I take this as an indication that introducing biogas plants in LMI countries could potentially be quite an effective “impact entrepreneurship” project, as it also might be increasing wellbeing via less air pollution and cheaper electricity. Speculatively, reducing the volume of waste lying around could also reduce the risk of pandemics, as trash can form a breeding ground for some diseases-carrying species.

A Tearfund report from 2019 suggests 400,000 to 1 million premature deaths per year are caused by waste mismanagement. The authors have considered 3 causes:

  1. Air pollution caused by open burning of waste (270-920 thousand deaths)
  2. Diarrhoeal disease (67-141 thousand deaths) caused by blocked drains, flooding, and other insanitary conditions arising from dumped waste
  3. Mosquito-borne disease (27-48 thousand deaths) including dengue and malaria caused by increased mosquito breeding grounds arising from dumped waste and flooding.

When it comes to waste harmful to animals and the environment, plastic waste comes up the most often. It’s estimated that plastic waste kills ~1 million seabirds and 100k marine mammals per year. From the point of view of the ecosystem, a particular problem comes from plastic waste which facilitates the proliferation of pathogens to coral reefs. However, these are widely discussed problems that don’t seem to have an easy fix.

Edit (02/24): In Not the End of the World, Hannah Ritchie suggests plastic pollution is actually the most tractable environmental problem, that could be solved tomorrow, if the Western leaders decided to finance waste infrastructure in developing countries. Most of ocean plastic pollution comes from a handful of rivers in Asia. Since we have this kind of infrastructure in Europe and North America, our waste is only responsible for ~5 % of the ocean plastic (Our World in Data). Presumably, such infrastructure would also lay ground for reducing the harms coming from other waste.


Another way to find priority paths within waste pollution might be to track particularly dangerous waste mismanagement in LMICs. One area of concern here, suggested in a post by Sahron Nyaga, is e-waste. Most electronic components contain lead, cadmium or mercury, which are strongly neurotoxic. The rise of leaded gas in the US in the 60’s and 70’s caused an average drop in IQ of ~5 points and significantly increased crime, perhaps including an 8 times increase in the number of serial killers.

The closest thing to an estimate of the health impact of e-waste I could find came from a recent report by the WHO. The report estimates the number of people exposed to e-waste as the simple number of informal waste workers, which is ~15-20 million people. As various sources note, informal waste workers are often rural migrants with low qualification, who don’t have an alternative, so rather than banning such work, some experts push for a wider usage of personal protective equipment. Such interventions, rather than outright bans, also don’t incentivize illegal dumpsites from forming and don’t require governments.

I haven’t been able to find sources which would attempt to quantify the number of people affected by e-waste indirectly (e.g. by diffusion into drinking water), which makes me guess these effects are relatively small, i.e. commensurate with the attention the problem is getting by NGOs and governments. 

However, as Sahron mentions, 60-90% of e-waste is imported or dumped illegally, which suggests improving the monitoring of electronics transport could be one way to mitigate it. I suppose there's a tiny chance this could also improve AI governance, in case some of the materials became a limited resource.

Long-term resource depletion will be solved by AI

In 1968, Paul Ehrlich attracted the attention of global leaders by his predictions that in the next two decades, hundreds of millions would starve to death due to overpopulation. Despite this widespread pessimism, the crop yields in Africa have doubled around the same period (1965-1970), thanks to the green revolution. Perhaps we should expect technological progress to surprise us again.

We can expect that a lot of the engineering that comes into making new materials or food sources will take a while to be automated but also, if AI soon automates most mental work, significant manpower may shift to doing the physical experiments necessary to advance material science. Since there are market incentives in making resources abundant, we should expect the kind of work we might make on this problem will get radically accelerated with the advent of AGI. We have at least one theory how a resulting radical abundance might be possible, nanotechnology. If such technology is to emerge, sacrificing our current wealth to future generations by switching to a 100 % circular economy likely will not pay off. Both because such technology could likely utilize our current waste and because a future abundance would make suffering at this age more important.

Additionally, the case for the long term impact of circular economy seems low because of its low neglectedness. Recycling & reuse should be motivated by market incentives, but also receive significant attention from policymakers - for instance, the EU is continually implementing solutions to the problem, China’s ban on plastic waste import has been praised by environmentalists and some NGOs are promoting similar bans worldwide.

  1. ^

    Some of the most extreme models like Swiss Re estimate that a warming of 2.5°C by 2050 would cause Africa and Asia to experience a counterfactual GDP decrease of ~21 %. Although this would reflect an awful hit, even such a hit wouldn't take us "far back", it corresponds to taking away India’s five years of progress and moving it from 2019 to 2014. Nevertheless, the Swiss Re report seems to fall among untransparent, top-down econometric models that Halstead criticizes.

  2. ^

    The theses Meier attacks are that biodiversity is the only thing that sustains life and that our existence depends on "sustaining of life (just) as we know it right now in the early twenty-first century".

  3. ^

    This label sounds only a little too extreme. 4.3 % of the Red List are critically endangered (>50% probability of going extinct in the wild within 10 years); 8.1% are endangered (>20 % p_extinction within 20 years); and 8.5 % are vulnerable (>10% p_extinction within 100 years). Source: Borgelt et al.'s supplementary materials + probability anchors from Red List definitions

  4. ^

    The first chapters of The Secret of Our Success seem like a good treatise on this question: On one hand European and American explorers who became stranded in nature usually died, unless they met indigenous locals, who taught them the culturally transmitted know-how of survival.  On the other hand, human spread across all climates is quite unique, possibly owing to our ability to accelerate our evolution by experimenting & copying successful behavior.
    It also doesn't seem that extinctions select for smaller species.





More posts like this

No comments on this post yet.
Be the first to respond.
Curated and popular this week
Relevant opportunities