The great energy descent (short version) - An important thing EA might have missed

tl;dr. We’ll probably have less and less energy in the coming decades, and should prepare for it. Here are the main takeaways:

  • Fossil fuels are finite and will peak in the coming decades

  • Renewables and nuclear rely on them to be built, depend on finite metals, or have limited growth potential, so they’ll have trouble scaling up

    • [Edit: Comments also pointed out that batteries and solar panels are improving, which I mention in post 1, but scaling them up to replace the entire fossil-powered infrastructure will take a lot of time, among other problems]

  • As economic growth relies on more and more energy, this would trigger a long-term recession, with many associated risks—possibly a crash of the economic system

  • Food production could decline seriously, and trade could be disrupted

  • We will probably not colonize the galaxy

  • However, this may also reduce our risk of going extinct from man-made causes, like AGI, climate change or bio-risks

[Edit: This post won a prize in the Red Teaming Effective Altruism Contest! This is appreciated. I’ll try to integrate the feedback it provides, so I shall make a follow-up post addressing the common counters given against claims of energy depletion—and explain why I’m still worried about this topic in spite of that]

[Edit: This post is apparently the “Most Underrated EA Forum Post in 2022”. I should admit that this was not among the things I expected.]

This post is a short version of 3 posts I wrote on the problem of energy depletion. These posts are organized like this:

1 - Energy, why it is the most important part of any society, the decline of fossil fuels and why alternative energy sources will probably not scale up

2 - Consequences: The role of investment, impact on economic growth and systemic risks. Plus, what that means for EA causes.

3 - What we can do, what we can’t do, and why few people really anticipate this problem

The claims below may sound very strong, but the data and full reasoning behind it can be found in these posts. You can check post 1 here (I advise reading them in order). I could not fit everything into these posts, so I also wrote a 140-pages-long Google Docs that goes into more detail about pretty much everything. If you have any objection in mind, I probably adressed it in these full posts or this doc.

[Edit: French president Emmanuel Macron, a neoliberal who has been a long-time skeptic of limits to growth and defender of technological progress as a solution, just said at the end of August that “We are living the end of what could have seemed an era of abundance… The end of the abundance of products, of technologies that seemed always available, [...] of land and materials”. This is a very strong message for a G7 president, which seems to indicate that we’re closer to limits to growth than we though. You can keep that in mind. More details here.]

The issue of energy depletion

The industrial civilization that we live in depends on a huge supply of affordable energy. Most of its components are very energy-intensive: food production, transportation, manufacturing of goods, extraction of resources… even the health sector. Most of this energy, even today, is provided by fossil fuels, especially oil, which is used in the fabrication or transport of almost all goods. However, these fuels are finite, which means that unless an alternative exists, our current growth-based industrial civilization cannot last.

The current viewpoint of EA on the topic, to the best of my knowledge, is that energy growth will continue for the coming decades, with technology and innovation solving the issue. However, I did a considerable amount of research in the last few years on this topic, and I have found severe limits everywhere I looked. From what I’ve seen, an “energy descent” seems a more likely prospect.

This conclusion differs considerably from what we usually hear, so I tried to present here the strongest claims made by a number of experts on energy depletion and limits to growth, which is a field EA has less exposure to (we have more of these experts in France).

It felt appropriate to include this post in the Red Teaming EA contest, especially as it challenges many longtermist worldviews. The main idea challenged here is the assumption that past growth trends will continue forever, and that the future of humanity will be an industrial civilization similar to the current one (an “extended present”).

In this 80 000 Hours post about existential risks, they suggest that we should “improve our understanding of whether any kind of resource depletion currently poses an existential risk”. The research done here suggests this could be the case.

The decline of fossil fuels

I present here the following claims:

  • Oil production will decline in the short to medium term (before 2040, it’s even possible that we won’t reach the levels of November 2018 again). The situation is different from the last oil shocks, since most countries today see a decline in oil production. If we look at the top 3 producers, Saudi Arabia announced that it would peak in 2027, Russia that it had peaked in 2019, and many executives from US energy companies believe that US production has peaked.

  • The reason for that is that oil extraction proceeds according to the low-hanging-fruit principle: the highest-quality and easiest-to-get resources are usually harvested first, so that we are now left with the stuff that is harder and costlier to extract. This is a diminishing returns problem. This doesn’t mean we will “run out”, as there are still a lot of resources in the ground. But it means that they will be less and less affordable each year. Natural gas and coal should last a little longer but not that much.

  • There is a very tight link between energy production and GDP growth. Some studies point to a one-way causality, with a dependency ratio higher than that of labor or capital. This makes sense, as goods and services need energy and natural resources to be produced. The data so far seems to indicate that an absolute decoupling is very unlikely at a global level. All of this means that a long-term decline of GDP will probably happen, unless we find a way to ensure a continuous supply of affordable energy. In the worst but probable case, long-term decline of growth means that we would have to redefine what we expect of human potential.

Can renewables replace fossil fuels?

As fossil fuels availability declines, there will be more investment in alternative forms of energy. However, there is no convincing lead to solve this problem of affordable and abundant energy supply at the scale needed. Alternatives work well at small scale, but replacing the entire fossil fuels system is extremely challenging:

  • Oil, coal and gas are used at every step of the manufacturing of all alternative energy sources. They require long and complex supply chains that are currently very dependent on cheap transportation by trucks, planes and ships (98% of which require oil). There could be alternatives, but they imply significant losses in efficiency (like for hydrogen) or geographical limits (like biofuels). The same issues apply for high heat in manufacturing, necessary for steel and concrete, and for plastic manufacturing, which are both very hard to substitute as well.

  • Metals are required for solar, wind, nuclear, batteries and electric cars, but the amount of metal you can extract depends on the amount of energy at your disposal. This is made worse by declining ore grades, which means that metal extraction is getting more and more energy intensive each year. Recycling is possible but never at 100%. Moreover, the supply of most metals is expected to be under tension at some point (except iron, aluminum and a few others), so substitution will be limited. Limits on energy will then probably mean limits on material extraction.

  • They might imply environmental consequences such as deforestation or competition with land (for biofuels), greenhouse gasses (for making gasoline or diesel with coal or gas), risk of nuclear proliferation (for nuclear power), local pollution and water depletion (for minerals extraction, e.g. for lithium)...

  • They need a lot of time to be deployed (building factories, smelting plants, mines, upgrading the electric grid…). Mines take 7-15 years to go from exploration to production, and nuclear plants need 10-15 years to be built. Energy transitions usually take between 50 and 70 years, so it’s hard to deploy them in a hurry.

With prices rising, won’t there be more investment and innovation?

To compensate for these limitations, I can see three possible options:

  • 1) Building an entirely renewable energy system that would be much bigger than the current fossil-based one, in order to compensate for losses in efficiency and the needs for storage. However, societies would have to allocate more and more of their resources in energy production every year, for the same amount of energy. This means that they can allocate less resources for other things like food, heating, transport, infrastructure or economic growth.

    • Moreover, prices will rise, but they cannot rise forever: if something gets too expensive, people simply cannot afford it. Some studies point out that when energy prices reach about 10% of GDP, a recession ensues. All of this puts a limit on energy investment.

  • 2) Doubling down on energy efficiency. However, efficiency cannot grow indefinitely, and the growth rate of efficiency declined in the last 50 years.

  • 3) Inventing an entirely new energy system. Many technologies exist or can be imagined, like fusion. However, we should also focus on the following questions: Can these technologies be developed and deployed fast enough? Are they truly sustainable, or do they depend on finite materials? If they are much more complicated and costly to develop, will they be attractive for investors?

Given these limitations, I think the most likely scenario is that, at a global level, we’ll have a decline in energy availability in the coming decades.

Of course, this is a complicated topic where it’s easy to get to wildly different conclusions, because our assumptions on how the economy and society work matter a lot. I had a lot of (great) discussions from ALLFED’s cofounder Dave Denkenberger, but we still do not agree on some key points. He tends to think that if a solution has been commercialized, then it will be scaled up to prevent energy descent because there will be enormous economic incentive to do so. I tend to think that there will be a limit on investment capacity because prices cannot rise too high, especially on things that require long-term investment like smelters or nuclear plants or mines.

Consequences on the economy and society

What would a decrease in energy availability imply? As oil is used in the transportation or manufacturing of 99% of products, this would mean that about every item gets less affordable. In the short term, our energy intensive and highly distributed economy would be vulnerable to an energy decline. Those shocks would translate into a more severe social and economic crisis, as the financial system isn’t geared to sustain a forced degrowth. This happened in the 1970s, but also in 2008, when the price of oil went from $20 to $140 in the span of 6 years. What follows after that is more speculative, however.

Having less and less energy each year would have major implications on several key topics:

  • Food is currently highly dependent on fossil fuels for transportation, fertilizers, pesticides and machinery. In the US, 10 calories of fossil fuels are used for 1 calorie of food. An estimated 4 billion people are alive today because of the surplus provided by natural gas fertilizers.

  • Transportation over long distances would be much more difficult. About every item around you has been moved by a diesel-powered truck or ship at some point. These liquid fuels are hard to replace, so trade would be more difficult over time. Countries that rely on international trade for their basic needs (food, heating, water, medicine), which is most countries, would be very vulnerable in case of a trade disruption.

  • Materials: A society with less energy would also have less and less cement, steel, plastics… Metals and materials would be harder to mine, extract, refine and transport. This means less goods and services of all kinds. This could trigger a feedback loop: less energy means less metals, but alternative energy sources rely on a larger amount of metals, meaning less metals to produce energy…

Another worry is that these issues might not arise gradually, but might come in the form of sudden shocks, too quickly for adaptation. For instance, a somewhat likely outcome is that financial institutions, deprived of growth, might crash down. Supply chains could be disrupted in such an event. Other events could also play in, like a country deciding unilaterally to keep its resources for itself. When crises arise, trust in the future tends to be lower, making cooperation and long-term planning harder. There are different ways to adapt to this problem, but leaders could adopt short-termist thinking, aiming for simple but suboptimal solutions. All of this would lead to greater political instability: agricultural disruption tends to cause civil conflict, which in turn causes interstate conflicts, especially in a period of perceived decline and rising inequality.

I do not have the skills to make cost-effectiveness estimates, but the scale of people affected would be great (most people on Earth, really). I personally assign a high probability (>85%) to an energy descent in the next decades. I also think a long-term recession is highly probable (a “Mad Max” scenario less so, but still possible). This is a topic for future research.

How this changes the future of EA

This could have very large implications for EA, and longtermism in particular, over some key points:

It’s possible that the idea of near infinite expansion of humanity with space colonization may not be an option. The articles I’ve seen on that point out that centuries of energy growth and technological progress would be necessary to solve the challenges linked to get to other stars—which appears unlikely. Indeed, the whole process of colonizing a new planet requires a lot of materials and energy. This could be a possible explanation for the Fermi paradox.

These considerations could significantly affect AI development, and there are some scenarios where the consequences of the energy descent could prevent its development. Preventing natural extinction risk, like asteroids, might not be possible, but man-made extinction risks are only a problem in a very energy-intensive society. I tend to think that finding ways to obtain even more energy would actually increase our probability of extinction from AGI, massive climate damage or bio risks, since we’d have more time to deploy them. Richard Heinberg makes the case that we are overpowered, having so much power (=energy) that we are able to wipe ourselves out accidentally. In such a case, I’m not sure a decline in energy availability should be prevented at all costs. The total value of the future would be lower than that in space colonization scenarios, but I don’t think the latter is a realistic option. This scenario might actually improve our future prospects compared to the median case.

In the long run, a major transformation of our society is to be expected with the plummeting of energy affordability. To make this transition as painless as possible, more work should be done to make accurate projections, and determine what futures are possible or not. EA forecasters could make a valuable contribution by producing forecasts that consider all facets of this issue carefully. A forecasting tournament on important indicators could be a good way to kickstart this. Even more ambitiously, there could also be forecasts on the probability of an “energy descent” scenario, once precisely defined.

Actions we could take

Many actions can be taken, and are proposed in post 3, but I’m not sure about which are the most cost-effective. These actions include:

  • Developing ways of making food that are less energy intensive (like ALLFED does).

  • Help make territories less dependent on external inputs for basic needs. Supporting policies that develop food autonomy and resilience, in order to be better able to accommodate a sudden disruption in the economy or supply chains.

  • Help develop the new political, social and economic systems that could be useful in the future, for when they might be needed. New worldviews will be needed, we can also help here.

  • Developing technologies that are more resilient, less complex, less energy-hungry, and easier to produce and repair.

  • Find ways to improve social cohesion, cooperation and decision-making.

Conclusion

It helps to look at this problem from a wider lens: we found a very large battery of stored energy in the ground, in the form of fossil fuels, and as it depletes, we’ll have less of it. This is an uneasy topic that challenges our worldviews deeply, but any projection into the future would be incomplete if it doesn’t take this topic into account. There is a lot we can do to prepare for a lower-energy future. If having forever more material goods is not an option, then we’ll have to learn how to live happily in a constrained world, in a truly sustainable way.

[Edit: It appears that many people stick to reading this post, which I understand, as the others are long- but I have to remind that this post is the most unconvicing one as it is only a summary version without much data.]

You probably have many objections in mind, which is normal, most of what we hear on this topic goes the other way. I highly recommend checking the full version, then, which is much better. It adresses stuff like why the recent improvements in renewables are not enough—same for efficiency.

The full version provides sources, data and graphs, and details why this is neglected, why energy transition models do not include what of what I’ve exposed above, and why the relationship between the economy and energy is so strong (there’s probably a bidirectional causality in there).

It is still somewhat summarized, though—the long version is the 140-pages additional doc.