silhouette of fireman holding hose

Exaggerating the risks (Part 5: Climate wrap-up)

As the Sun slowly grows brighter over its main sequence lifetime, habitability on Earth’s surface will eventually become threatened probably leading to moist and then runaway greenhouse climates … Nonetheless, our results imply that Earth’s climate may remain safe against both water loss and thermal runaway limits for at least another 1.5 billion years and probably for much longer.

Wolf and Toon, “Delayed onset of runaway and moist greenhouse climates for Earth
Listen to this post

1. Recap

This is Part 5 of the series Exaggerating the risks. In this series, I look at some places where leading estimates of existential risk look to have been exaggerated.

Part 1 introduced the series. Part 2 looked at Toby Ord’s claim that there is a 1/1,000 chance of irreversible existential catastrophe from climate change by 2100. We saw in Part 2 that Ord gives only very thin support to this claim, but to give Ord the benefit of the doubt, we took a look at all plausible ways in which climate change could lead to existential catastrophe in the near future.

Part 3 and Part 4 of this series used a report by John Halstead to look at a number of plausible extinction mechanisms, and found them unconvincing. We also saw in Part 4 that the paleoclimate data which Ord uses to ground his risk claims does not support a high level of existential risk from climate change in the near future.

Today, I want to wrap up the discussion of climate risk by drawing on Halstead’s treatment of two remaining risks: moist and runaway greenhouse effects, and interstate conflict. Then I will draw general lessons from Parts 2-5 and map out the future of this series.

2. Moist and runaway greenhouse effects

As the world warms, evaporating water vapor will be entrapped within the atmosphere, leading to additional warming. If the earth reaches high temperatures (perhaps around 67°C) we will experience a moist greenhouse effect in which, over millions of years, the oceans boil off into space. If matters become very severe, we could experience a runaway greenhouse effect in which the planet warms rapidly to over a thousand degrees Celsius. This may well have happened on Venus.

These figures suggest the natural response to fears of moist and runaway greenhouse effects: it is highly implausible that the earth would soon undergo the over 50°C of warming needed to trigger a moist greenhouse effect, and if this were to happen we would be long dead of other causes before a moist greenhouse were triggered. Many of Halstead’s responses begin from the implausibility of a moist greenhouse effect being triggered.

First, Halstead reviews the conditions across a variety of studies that would be required for a moist greenhouse effect to be triggered. Here are, for brevity, the first five:

John Halstead, “Climate change and longtermism

However, Halstead argues, depending on how we measure the amount of accessible fossil fuels in the world, burning literally all of the world’s accessible fuels and doing nothing to mitigate our folly would bring atmospheric CO2 concentration to somewhere between 1,600ppm and a bit over 5,000ppm. This means that even in the very unlikely scenario that humans exhaust our entire stock of fossil fuels and do nothing to offset these emissions, a moist greenhouse effect remains highly improbable.

Halstead also reminds us (as we saw in Part 4) that the earth has already been at least 17°C hotter than preindustrial levels. This means that if we think lower levels of warming could risk triggering a moist or runaway greenhouse effect, we will have a difficult time explaining how these effects are meant to come about now when they were not triggered in previous, warmer climates.

Finally, Halstead reminds us, even if we did transition into a moist greenhouse effect it would take millions of years for the oceans to boil off. This means that humans would have ample time to reverse the effects of warming unless we entered a runaway greenhouse regime. We could reverse warming using complicated methods, such as decarbonization. But on this timescale, we could also reverse warming through simpler methods, in large part by simply reducing our emissions and waiting for previously emitted CO2 to dissipate.

Summing up, there is very little reason to believe that the earth is likely to experience a moist or runaway greenhouse effect any time soon. If this did happen, we would probably be dead long before it begun, rendering the appeal to moist greenhouse effects a moot point. And if we somehow managed to survive until the beginning of a moist greenhouse effect, we might well be able to reverse it.

3. Interstate conflict

One final way in which climate change could lead to existential catastrophe is by increasing the risk of interstate conflict. Climate change will exacerbate many forms of scarcity, deprivation, and desperation that can lead to instability both within and across nations. That instability can lead to war, which could pose an existential threat.

Here Halstead begins with an important reminder: most effective altruists, including Ord, think that even extreme forms of warfare such as all-out nuclear war are unlikely to produce human extinction. That is not to say that wars would not be brutal. But we should not be quick to infer from even the bleakest interstate war to the inevitability of human extinction or the permanent curtailment of human potential for desirable future development. Humans are a resilient species. We have survived many wars in the past, and we will likely survive many wars in the future.

Beyond this, Halstead notes, the overall trend in recent history has been a sharp decline of interstate conflict. For example, Halstead cites a report by Our World in Data which charts the recent history of great power conflict:

The last few centuries have brought a sharp decline of great power conflict, from near-incessant conflict to very low conflict.

Even in more recent history, despite increases in population and destructive technology, deaths in interstate conflicts have declined sharply:

The suggestion implicit in these graphs has been echoed by most recent scholars: interstate conflict is declining in severity, and great power conflict is declining even more rapidly. This means that even if other factors, such as climate change, tend to exacerbate conflict, climate change will have to contend with countervailing factors that have significantly reduced conflict in recent history, and may continue to reduce future conflict.

Now it is true that some scholars think climate change may exacerbate future conflict, although this finding is controversial (See Burke et al. 2015, Koubi 2019, Saleyhan 2008). However, almost all scholars agree that climate change is far from the most influential factor in driving conflict. For example, a recent expert elicitation in Nature found that experts ranked climate change as among the least influential factors driving interstate conflict.

Readers who find themselves troubled by the association between climate change and conflict might ask themselves whether they are also troubled by the higher associations between quantities such as intergroup inequality and conflict. It may also be worth considering whether factors such as increasing socioeconomic development, decreasing corruption, and reduced natural resource dependency may serve to offset or even reverse the effects of climate change on interstate conflict.

Finally, Halstead considers the association between climate change and great power war, which may pose the greatest risk of existential catastrophe. Halstead examines four mechanisms by which climate change could lead to an increase in great power war:

  • Water scarcity: Scarce water supplies could become a source of conflict.
  • Economic costs of climate change: Economic upheaval produced by climate change could increase the risk of conflict.
  • Civil conflict: Intrastate conflicts produced by climate change could spill over into interstate conflicts.
  • Mass migration: Mass migration away from low- and middle-income countries could produce political instability in great powers, raising the risk of conflict.

In each case, Halstead argues, the risk increase is not so great.

For example, on water resources Halstead cites a study by Aaron Wolf and colleagues (2003) which categorizes interstate water disputes and cooperation since 1948 on the Basins at Risk scale, ranging from -7 (unification into one nation) to +7 (formal war). Out of many hundreds of incidents, Wolf and colleagues find not a single instance of formal war.

This isn’t to say that war over water is impossible, but it should go some way towards allaying fears that war over water is inevitable.

Likewise, on the relationship between war and economic distress, scholarly opinion is decidedly mixed. Some think there is an association to be found, whereas others are more skeptical. Summarizing the literature, Halstead cites a report by the political scientist Greg Cashman on the causes of interstate conflict:

Probably the most we can say about the business cycle is that it may play a role in the development of war, but its effects are less than clear. Some wars have broken out in hard times, others have occurred in good times. Neither economic weakness nor prosperity seem to prevent war.

Greg Cashman, What causes war? An introduction to theories of interstate conflict

Halstead also reviews the two remaining mechanisms that could link climate change to great power conflict, but I think we have seen enough.

Summing up, it is far from clear that great power conflict would lead to existential catastrophe. Most scholars see only a very weak relationship between climate change and interstate conflict, and take other factors to be much more influential. Many of those factors are cited in a trend of rapidly declining great power conflict, as well as rapidly declining conflict deaths. Moreover, there are no clear mechanisms through which climate change could substantially exacerbate the risk of great power conflict.

None of this is meant to suggest that climate change will have no effect on the risk of existential catastrophe through great power conflict. But neither would we be warranted in taking this data to suggest a substantial chance of existential catastrophe from climate-induced great power conflict, or any other form of climate-induced interstate conflict, any time soon.

4. Preliminary lessons

In this series, we looked at Toby Ord’s claim that climate change poses a 1/1,000 chance of irreversible existential catastrophe by 2100. We used Halstead’s report to examine a range of mechanisms by which near-term climate change could lead to existential catastrophe, including crop failure, heat stress, rising sea levels, climate tipping points, moist and runaway greenhouse effects, and interstate conflict. In each case, we found the evidence unable to support anything like a 1/1,000 chance of irreversible existential catastrophe by 2100.

Suppose that is right. What follows? At least three lessons are suggested by this discussion.

First, risk estimates can be inflated by orders of magnitude. For example, we saw in Part 2 that Ord assigns a 1/1,000 chance that climate change will lead to irreversible existential catastrophe by 2100. However, we saw that Halstead struggles to get the risk above 1/100,000, not merely by 2100, but in fact over all time.

If this is right, then we need to approach remaining risk estimates with a degree of skepticism and caution. If even leaders within the effective altruism movement, speaking in books widely cited as authoritative references on existential risk, can be off by orders of magnitude in their estimates, then we should not be too hasty to accept risk estimates provided elsewhere by effective altruists.

Second, the evidential basis for existential risk estimates is sometimes very slim. We saw in Part 2 that Toby Ord’s estimate of climate risk is based on at most a few sentences within a long discussion. This carries a few dangers.

First, slim evidential bases are often wrong, or at least wrongly interpreted. For example, we saw in Part 3 that the paleoclimate data which Ord appeals to should not be interpreted as supporting high estimates of climate risk. If anything, a full look at paleoclimate data suggests the opposite conclusion.

Second, slim evidential bases leave so much out that it is difficult to know what to make of some existential risk estimates. We spent three parts of this series tracking down all possible bases for Ord’s climate risk estimates, precisely because Ord gives us so little evidence for his estimates that it is hard to know what he was driving at. It should not be the burden of skeptics about existential risk to reconstruct fragments cast out in support of risk estimates. Risk estimates need to speak for themselves.

Finally, we saw throughout this series what I have called a regression to the inscrutable. Effective altruists begin by chronicling risks that are relatively tractable using ordinary scientific methods, such as crop failure, heat stress, or flooding. Finding these risks to be slight, effective altruists increasingly place most of their confidence in esoteric risks such as tipping cascades or runaway greenhouse effects. These risks are distinguished by the fact that they are increasingly inaccessible to our best scientific methods.

Now of course, it could turn out that most existential risks are in fact highly inscrutable risks. But if it instead turns out that across many different risks, effective altruists invest most of their confidence in relatively inscrutable risks, then a different explanation suggests itself. In that case, it would appear that effective altruists may be using the inscrutability of risks as a way to mask insufficient evidence for high risk estimates, as well as failing to update correctly on the consistent smallness of scrutable risks.

Effective altruists, including Ord, consistently invest low credence in natural risks such as supervolcanoes, asteroids, and naturally occurring pandemics. These risks are relatively scrutable, and scientific investigation has shown them to be low.

Likewise, effective altruists often invest low credence in risks such as nuclear war that, while initially assumed to be high, were later shown by scientific research to pose a minimal threat of outright human extinction.

Increasingly, effective altruists invest most of their confidence in relatively inscrutable risks such as bioterrorism and risks from misaligned artificial intelligence. Indeed, Ord puts the combined existential risk from these two sources at over 13% by 2100. As the focus of this series shifts to more inscrutable risks, readers would do well to bear in mind the failure of more scrutable risks, such as climate risk, to stand up to scientific scrutiny, as well as the tendency within more scrutable areas, such as climate risk, to concentrate risk estimates within the least scrutable sources of risk.

5. The way forward

That is all I have to say about climate risk. My current plan is to move on to a discussion of some leading arguments for existential risk from misaligned artificial intelligence, perhaps pausing briefly along the way to discuss some historical claims about existential risk which now look to have been exaggerated.

Please do let me know what you think of this series, and what risks you are especially keen to hear about in future continuations. Comments are open.


2 responses to “Exaggerating the risks (Part 5: Climate wrap-up)”

  1. Amadeo Avatar

    I look forward to hearing your case against high AI risk, and your view on where the exaggeration of existential risks as a whole would leave the tractability of longtermism if it were true. (Many people seem to point to existential risks as the strongest evidence against cluelessness. If those are exaggerated by orders of magnitude, are longtermists left clueless again?)

    Also, I’m wondering why you don’t cross-post your posts to the EA Forum. I suspect this would increase both the number of readers and engagement by EAs with your arguments.

    1. David Thorstad Avatar

      Thanks Amadeo!

      I think it’s under-appreciated how many different arguments made by effective altruists tend to rely on the idea that AI poses a significant existential risk. In this regard, I’m happy to see that we agree on a leading strategy for resisting the cluelessness objection being to (a) point out some near-term existential risks and (b) point to AI risk in particular, as the best example of a strong and tractable risk. It’s certainly not the only move that can be made to resist cluelessness, but it is one of the most common moves.

      I’m always happy to see discussion of my work in other venues. I would be very happy if someone wanted to cross-post my blog posts or some of my academic papers on the EA Forum.

Leave a Reply