Exaggerating the risks (Part 2: Ord on climate risk)

There is no evidence of climate change scenarios that would render human beings extinct.

Michael Mann, cited in “Could climate science make humans go extinct?

Sometimes, effective altruists give high estimates of the existential risks facing humanity today. In his bestselling book The Precipice, Toby Ord puts the risk of existential catastrophe by 2100 at “one in six: Russian roulette”.

That is a very large number. What does the evidence say? Let’s look at the case of climate risk.

1. Why climate change does not pose a significant existential risk

Ord gives a 1/1,000 chance that climate change will lead by 2100 to irreversible existential catastrophe. That is a frightening number.

But the odd thing about Ord’s discussion is that almost the entirety of the text is devoted to a clear and well-evidenced discussion of why climate change is unlikely to lead to existential catastrophe any time soon.

How could climate change produce an existential catastrophe? Ord argues against six routes to existential catastrophe. Like many of Ord’s writings, this discussion is admirably clear, well-researched, and persuasive.

First, climate change could trigger a runaway greenhouse effect. Emissions would trigger water evaporation, leading to a buildup of water vapor in the atmosphere. That, in turn, would cause more evaporation, until the oceans had boiled off. But, Ord notes, while such an event has occurred on at least one planet, our best scientific models suggest that humanity is incapable of producing even a fraction of the emissions needed to trigger a runaway greenhouse effect on earth.

Second, we could trigger a moist greenhouse effect, a little cousin of a runaway greenhouse effect whose consequences would involve far less evaporation and warming. But again, Ord correctly holds, our best scientific evidence makes a moist greenhouse effect unlikely to happen any time soon, and models predict resulting temperature changes that would be survivable, though catastrophic. 

Third, there could be other feedback effects such as melting Arctic permafrost or release of methane from the deep ocean floor. But these are not likely to pose an existential threat this century. Ord notes that even under its “high emissions” scenario, the IPCC expects melting permafrost to contribute 0.3°C of warming by 2100. And although very little is known about the likelihood or scale of ocean methane release,  Ord gives no reason to suspect that this would be fatal.

Fourth, we can (and probably will) continue to burn fossil fuels. But, Ord notes, even the implausibly extreme scenario in which we burned literally all accessible fossil fuels, predicted warming would be in the range of 9 to 13 degrees Celsius. That is an astounding number, and certainly poses a catastrophic risk, but there is absolutely no reason to think it would not be survivable: many regions of the earth would remain habitable, above water, and friendly to agriculture.

Fifth, global warming could cause biodiversity loss and ecosystem collapse. But, Ord notes, the evidence linking climate change to biodiversity loss, let alone ecosystem collapse, is mixed. And Ord gives no reason to think that either scenario would be unsurvivable.

Finally, global warming will doubtless cause heat stress, rendering certain areas of the earth uninhabitable during the warmest months without air conditioning. But many areas of the earth will remain temperate, and the rest may be inhabitable with the use of air conditioning and other technologies.

So far, the entirety of Ord’s discussion has given us reasons not to take climate change to pose an existential threat to humanity. We have not been given a single reason to take climate change to pose an existential threat. I don’t mean that we have been given a bad reason. I mean that we have literally not been given a single reason.

On this basis, one might expect Ord to treat existential risk from climate change as negligible. But in fact, Ord gives an estimate of one in a thousand that climate change will produce an existential catastrophe by 2100. That is a shockingly high number. But what, exactly, lies behind it?

2. What might underly Ord’s risk estimate

The only sliver of an argument for existential risk from climate change is contained in a single paragraph. That paragraph begins as follows:

This [discussion] doesn’t rule out unknown mechanisms. We are considering large changes to the Earth that may even be unprecedented in size or speed. It wouldn’t be astonishing if that directly led to our permanent ruin.

And now we are surely going to get an argument for why we should be afraid of these unknown mechanisms? No. In fact, we get an argument against that fear.

The best argument against such unknown mechanisms is probably that the PETM [Paleocene–Eocene Thermal Maximum] did not lead to a mass extinction, despite temperatures rapidly rising about 5 degrees Celsius, to reach a level of 14 degrees Celsius above pre-industrial temperatures.

So far, we have been given nothing to fear. Ord does offer a reason not update too strongly on the PETM:

But this is tempered by the imprecision of paleoclimate data, the sparsity of the fossil record, the smaller size of mammals at the time (making them more heat-tolerant), and a reluctance to rely on a single example.

Fair enough! But what is the positive reason to be afraid of existential risks from climate change? Finally, finally we are given a sliver of an argument:

Most importantly, anthropogenic warming could be over a hundred times faster than warming during the PETM, and rapid warming has been suggested as a contributing factor in the end-Permian mass extinction, in which 96 percent of species went extinct.

Is that the driver behind the view that we have one chance in a thousand of existential catastrophe from climate change by 2100? Surely not. No one would leap to such a strong view on the basis of this single sentence.

But then, what exactly is the argument that climate change poses a near-term existential risk? Could it be that there is no argument? That we have as yet been offered no solid, rational, evidential basis for fear?

“That’s not fair”, you say. “Ord wasn’t claiming to have given us an argument for his existential risk numbers. They were just subjective reports of Ord’s own views, which were never intended to be treated as statements of fact or as reflections of printed arguments.”


Have I been fair to Ord? Are there other reasons to be afraid of existential risk from climate change? Let me know.


7 responses to “Exaggerating the risks (Part 2: Ord on climate risk)”

  1. loophole Avatar

    If you think 1/1,000 chance of extinction by climate change is a “strong view”, what do you think is a reasonable prior on extinction risk?

    Ord calculates priors on extinction risk per century using the age of humanity and survival of related species, and he comes out with 1/2,000-1/100,000. Seems to me (to make up some numbers) that 1/1,000 would be a reasonable prior for extinction risk from anthropogenic climate change, and the fact that we’ve ruled out the mechanisms we know of might reduce that by 10-100x—not all that far from Ord’s estimate.

    Seems like you’re saying that in cases involving a lot of uncertainty, we should assume there’s nothing to fear. Why would that be?

    1. David Thorstad Avatar

      Thanks loophole! I guess I would want to say a few things here. The first is that we need to sharply distinguish prior beliefs about extinction risk from prior beliefs about extinction risk *from climate change*. It’s quite tricky to infer from prior beliefs about extinction risk to prior beliefs about extinction risk from climate change, since any number of things besides climate change could raise a risk of extinction.

      The second is that we need to be careful about making inferences about human extinction from the extinction rates of past nonhuman species. That’s because humans can do much more to adapt to threats that tend to kill other species. For example, in the face of climate change we can fight heat stress with air conditioning. We can fight the death of plants by planting crops, even genetically-engineered crops. And we can even take very extreme actions like working towards decarbonization (or just sensible actions like reducing fossil fuel use). So humans might be in a better position to survive threats such as climate change than other species would be.

      More broadly, I think you are quite right that we need to look at mechanisms. My aim in this post, and the next two posts in this series, is to look at every plausible mechanism by which climate change could lead irreversibly towards existential catastrophe by 2100. I’ll argue that each mechanism is very unlikely to lead to existential catastrophe by 2100. If no such mechanism is likely to lead to existential catastrophe by 2100, then we shouldn’t place much credence in the idea that climate change will lead to existential catastrophe by 2100.

      1. loophole Avatar

        Thanks! I think we agree that not much is known about, e.g., methane stores in the ocean, so our risk estimates will be very sensitive to priors. Sounds like for you to be worried about existential risk from climate change, you’d need to see an end-to-end mechanistic story in which climate change leads to extinction; that makes me think you put a very low prior probability on all-cause extinction. If so I’d love to hear more, because that seems to be our main disagreement.

        My own reasoning about existential risk this century is something like this: (1) Prior from reference-class forecasting: 1/10,000; (2) Big adjustment down for human adaptability, as you mention, but also big adjustment up for technological capability: ??? but let’s say 1/100. This leaves lots of extinction probability-mass to go around, so it seems reasonable that some of it would land in the “climate change” bucket.

        My first comment wrongly suggested that your disagreement was probably with (1), but I imagine it’s in fact with (2).

    2. David Thorstad Avatar

      Thanks loophole. ((I’m replying here to your second comment because I haven’t yet figured out how to get nested comments to display more than 3 levels deep without breaking the layout. I know that EAs are a tech savvy bunch, so if you or anyone else knows how to make deeply nested comments work on wordpress, please let me know!)).

      What a fortuitous coincidence – another commentator on this thread (Gavin) just mentioned deep ocean methane too. I’ll talk a bit about methane later in this series in my discussion of tipping points, but in general I think Halstead’s discussion here is spot on. And in particular, I think Gavin’s summary of the evidence here is revealing: a lot of the force behind the worry about deep ocean methane relies on some rather fringe science.

      I do think that a good deal of my disagreement is with (2). I think that when a good deal of evidence is available (as it is in the case of climate risk by 2100), the role of priors should mostly wash out, and readers with any number of very different priors should generally tend to converge on the view that climate risk in this century is quite low.

      I do have some worries about (1). As I mentioned, I am not sure how much weight I would give to reference class forecasting as a means for estimating the survival chances of humans. That is because humans are radically different from all other species in the reference class in our technology, intelligence, and ability to foresee and adapt to risks. (We’re also, unfortunately, radically different in our ability to cause risks).

      More generally, I would tend to begin deliberation with a much lower credence in human extinction than many EAs seem to have. I think that some EA commentators are often rather afraid of what the future may have in store, and that they are afraid enough that when they are not able to find plausible mechanisms for human extinction, they settle on implausible mechanisms (sleeper viruses, rogue superintelligence) or invest high credence in unknown mechanisms (which, to me, is just a recipe for fear, since it is hard to talk someone out of their fears if they can’t tell you what they are afraid of).

      Another thing that I will try to stress in this series is that many previous fears of existential catastrophe have been substantially overblown – not merely in that they did not come to pass, but that in hindsight we can see that the estimates were exaggerated and driven by distorting factors such as the salience of present technological developments or fear of the unknown. I think that this history of inflated predictions of existential risk should give us general reason to discount some of the higher risk estimates that are being given today, insofar as those risk estimates may be driven by something like the salience of present technologies (AI, CRISPR …) or a fear of the unknown.

      1. loophole Avatar

        Thanks David. All fair points and I’m interested to read more!

  2. Gavin Avatar

    As chance would have it I recently found an actual reason. Not a great one, but a supposedly precedented one.


    1. David Thorstad Avatar

      Ah, nice point. As luck will have it, another commentator (loophole) just mentioned methane too.

      I think that your analysis is spot on, and illustrates a few general lessons that I’d like to stress in this series.

      (1) Science matters. The last thing we want to do is to make empirical predictions based on fringe science, or no science at all. Ward is not speaking for the scientific community (indeed, Ward is a paleontologist, not a climate scientist). And as you note, there are “hallmarks of crankery in the book – most of the citations of the book are unscientific or pseudoscientific”. I think it is a mistake to take risks to be high on the basis of crank science, and I think you make this point nicely.

      (2) In thinking about existential risk, there is what I call a regression to the inscrutable. We start with the most empirically and scientifically tractable risks, such as asteroids, nuclear war and climate change. Then we crunch the numbers and realize these things aren’t actually very risky. Then we turn to things that are harder to analyze, such as risks from deep ocean methane. Then we crunch the numbers and realize these things aren’t very risky either. And we continue retreating into deeper and less scrutable risks until we end up in a place where almost all risk is hypothesized to come from very difficult to analyze scenarios (killer sleeper viruses; rogue superintelligence). When a community finds itself constantly supporting its positions by putting increasingly high confidence in a range of ever more inscrutable scenarios, it begins to seem to many outside observers that evidence is no longer playing a strong driving role in the debate. What is really happening is that many effective altruists are quite fearful of existential catastrophe, and willing to stick to that fear by moving it further and further away from the reach of counterevidence. It is quite hard to talk someone out of such fears because, by construction, they are highly resistant to counterevidence. But it is also very hard for others, who are not so fearful, to see the need for fear.

Leave a Reply

%d bloggers like this: