One month ago, I wrote a blog post in which I made the following claim:

Ukraine's ability to resist a Russian invasion by itself is even less than Poland's ability to resist a German invasion in 1939, and I don't think anyone doubted this.

I'm leaving this claim up in the original post for the record, but here I want to think about what exactly went wrong in my model of the war. I think the broader point I was making in the linked blog post is accurate, but I was overconfident here in a way I should not have been and this is a retrospective of me trying to explain why that happened.

When I wrote this blog post, I was fairly confident (over 80%) that the war would end up going well for Russia. It hasn't, and I'm quite surprised by how poorly the Russians have ended up performing. Looking at some past wars through this lens, I think wars going in unexpected ways is actually normal throughout history, and I should have been a lot less confident in my evaluation of how the war would turn out.

I think the main source of error in my model here was that I didn't take into account the degree to which Germany mobilized against Poland. Germany invaded Poland with an army of close to 2 million soldiers against Poland's initial defense force of 250 thousand. Poland later got this number up to around 800 thousand by mobilization, but the Germans still had a 2:1 numerical advantage.

In contrast, Russia has tried to invade Ukraine with an invasion force of around 200k soldiers, and they are likely outnumbered by the defenders now by at least 2:1. Defending is easier than attacking in war due to many reasons, so the size of the Russian invasion force should have led me to be more skeptical about this comparison with Poland.

Another reference class I had in mind was the 2003 US invasion of Iraq. Here the Iraqi defenders outnumbered the coalition forces 2:1 but were handily defeated anyway. I think as a comparison case it was definitely the right decision to keep this in mind, but I didn't properly take into account the degree to which war is an uncertain affair and so I updated too strongly on these few examples without appropriately weighing the counterexamples.

For instance, in 1940 Germany invaded France with an invasion force roughly of the same size as the one that they used in Operation Barbarossa. France fell in six weeks while the USSR eventually went on to mobilize more men and drive the Germans back. Furthermore, the reputation of the French army was in fact superior to that of the Soviet army at this time, and having defeated France in six weeks definitely emboldened German leadership to believe the USSR would similarly be easy to defeat. In July 1941, German field marshal Halder even went so far as to claim that "the war against Russia has already been won".

Could the rapid collapse of France in contrast with the prolonged resistance and eventual victory of the USSR have been foreseen in advance? I don't think so. There has been plenty of ex post storytelling about this, but I don't think it could have reasonably been foreseen at the time.

To a lesser extent, I was also taken in by the mystique and reputation of the Russian army despite believing that I was appropriately skeptical, a mistake I was far from alone in making. Before the war I gave a measly 5% chance to a Winter War scenario, and despite how poor this prediction ended up being I've actually been told that it was "prudent" simply because many others I know were even more overconfident than I was. An instance of this even happened on LessWrong itself, in lsusr's post about the invasion:

There is probably going to be a war. Ukraine is probably going to lose. The question is how much, how quickly and on what terms.

Eastern Ukraine is a flat plain contiguous with Russia. If you just look at troop counts then Ukraine would seem to have a chance against Russia. But Russia has superiority of aircraft and heavy weapons. Russia will conquer Eastern Ukraine. The Russian Armed Forces is among the three most capable militaries in the world. The Ukrainian military isn't.

I can understand anyone who made this mistake since I made it myself. I hope I learn from this mistake and I'm putting this post up both for the sake of transparency and so that others may learn from my mistakes instead of having to make similar ones themselves.

38

35 comments, sorted by Click to highlight new comments since: Today at 8:36 AM
New Comment

I wasn't surprised by Ukrainian resistance. In the article you link, I wrote "The Ukrainian government will fight a total war to defend its sovereignty." which was bet against in the comments. I was surprised by Russian incompetence. My model for why I was wrong goes like this.

  1. I wasn't paying attention to Russia's build-up on Ukraine's border. When Russia invaded, I quickly copied the predictions of Western experts on Russia.
  2. Western experts copied their predictions from public reports by US intelligence.
  3. US intelligence copied their predictions from secret reports stolen from Russia.
  4. Russian intelligence was garbage due to systemic corruption and the fact Putin didn't tell his lower echelons they'd be invading Ukraine for real.

Ironically, I wrote a story last year satirizing a world where every intelligence agency just steals each other's data.

That's fair. Personally I'm skeptical of any causal chain with too many moving parts like this because the probability of the whole chain holding up is the probability of each step going forward conditional on the previous steps and that product can get small very fast. Still, it's a funny explanation and deserves an upvote for that alone.

I was paying attention but for the reasons mentioned in the post (and possibly others, maybe groupthink?) I ended up being way more confident in Russian success than I should have been. I wasn't copying predictions from Western experts in any direct way, though some of their model of the world likely blended over into mine in a process of gradual osmosis.

I also wasn't surprised by Ukrainian resistance; I just thought that it would be more analogous to Polish resistance than Finnish resistance (in 1939) in how successful it would end up being.

I am generally pretty pleased with my model of the war. I have had angry conversations with "well-informed" people for years about overestimating the capabilities of the Chinese and the Russians. I'd often three-quarters-jokingly accuse them of being closet authoritarians with how much competence and sincerity they projected onto their opponents' armed forces and how much (accurate) cynicism they had about their own ranks. It has been deliciously satisfying to see some of them come around to my point of view over the course of the last month. 

My largest problem was underconfidence; I started out the war with this pervading thought pattern that, well, maybe they're going to come out and do something so crazy and ridiculous that I just can't even conceive of it because I'm so westernized. I was unwilling to put any LW street cred on the line until something like a week into the conflict. I missed a very real opportunity to be Mega Cool and publicize "75% chance of Ukrainian semi-victory" 24 hours in. 

I'm pretty bullish on what Chinese military capabilities will look like in a few decades. If you think Western observers are overestimating the capabilities of the Chinese then now is a great time to make a statement on the public record to that effect. It's not often we get good opportunities to make long-range falsifiable bets against mainstream beliefs about important issues.

(Technically your above comment constitutes such a statement.)

It's not often we get good opportunities to make long-range falsifiable bets against mainstream beliefs about important issues.

 

Financial markets are full of such opportunities. 

While I agree with a lot of what you said in the post - particularly about the respective capabilities of China and the USA in 2050, and the vulnerability of aircraft carriers, I think the current war should change your assessment of Taïwan's chances in a defensive War.

Why? Much of Ukraine's defensive capability comes from its supply lines from the West. If my 2050 analysis is correct, China could easily knock out ships from America and Japan. Taiwan would have to operate without resupply.

My sense of it is that wars not turning out as you expected should decrease your confidence in wars turning out as you expect in the future, just for outside view reasons.

Will be hard to go on public record in a satisfyingly concrete way without going through a wargame like you did, so I'm gonna plan on doing so in the next couple months. 

I'm bearish on Chinese military capabilities, but even more importantly I disagree on the direction relative power between China and the US is going. I think we have already passed peak-China and it's all downhill from here.

Why do you think that ? Simply through pure economic growth I'd expect them to continue improving...

Overpopulation. I believe we live in a zero-sum world where labor is highly replaceable but resources are limited. China's has too many people to begin with; its declining birth rate helps to relieve population pressure in the long term but in the mid term the bulk of their population is slowly entering retirement age, creating a huge additional huge burden.

I believe we live in a zero-sum world where labor is highly replaceable but resources are limited.

How is this possible if world population has been soaring along with per capita income/living standards for the last three hundred years?

There is no contradiction here. The soaring living standards over the last few centuries is the result of rapid technological advance, which I don't dispute. What I mean by "overpopulation" is that we would be even better off, have a higher per capita income, if there was less people at the given tech level.

Imagine what would happen if we dialed world population to extremes: 

At one end, you would be the only person in the world. Theoretically, all the world's resources belong to you, making you very rich. But in practice, you will live an impoverished and short life, because you can't run a modern economy all by yourself to make use of those resources.

At the other end, the world is populated by 200 billion people and everyone is on the brink of starvation, surviving in urban slums on a diet of insects and algae.

Somewhere in the middle, there is an optimum level of population that maximizes wealth per capita. This dynamic is a version of the Laffer curve, which tends to arise whenever two opposing factors are at play: For government tax revenue, it is the diminishing incentive to create wealth vs. the expanding fraction of wealth captured by the state. For aerodynamic resistance in rocketry, it is the decreasing air density vs. the increasing speed of the vehicle. For world's per capita income, it is the shrinking share of resources per person vs. the soaring economy of scale in production.

Notice that this optimum is decoupled from the absolute per capita income. The world can simultaneously be increasingly overpopulated and improving in terms of living standard because technological dividend is outpacing population growth. It could even be the case that the world is becoming more overpopulated despite population remaining constant, due to automation eating into the marginal utility of labor.

In a further step, I believe there is overpopulation (i.e. oversupply of labor) at all levels, from unskilled labor all the way to the Einsteins of the world. I think the world's best scientists/engineers are mostly positional goods, similar to the world's best athletes - in a counterfactual world where they never existed, things would be chugging along at the same pace. This is why I'm not worried at all about the staggering number of university graduates and engineers China is producing each year - all this does is intensify internal competition.

For this reason, I see the post-WW2 globalization of economy  as largely a redistribution scheme. I still think it was worthwhile, but only on moral grounds rather than economic self-interest of the powers that promoted globalization. Urbanization and rising living standards are among the most effective methods of hampering population growth, alleviating the problem at the source. Globalization has also been a promoter of peace, thus avoiding burning resources for nothing in wars.

The soaring living standards over the last few centuries is the result of rapid technological advance, which I don't dispute. What I mean by "overpopulation" is that we would be even better off, have a higher per capita income, if there was less people at the given tech level.

We wouldn't have the tech level that we have now without billions of people, but I'll save that argument for later. 

Here is a more relevant critique of your position: Notice however that the issue here is per capita income. The relevant metric for wartime capability is a nation's total productive capacity. Total productive capacity of a nation does increase with population and we can prove this with a simple thought experiment.

New babies are not born with a ticket that entitles them to a percent share of world GDP or land or iron ore. In practice they get a lot of support from their parents if they are to survive, but that's not really relevant for the purposes of this thought experiment because we can imagine them now at 18 years old and kicked out of the house with no money. Since on average regular people, even people who start out without assets, still find ways to sell their labor on the market for more than it takes for them to survive, we can assume that, on the margin, more workers contribute to China's economic production. Even if your hypothesis is correct and their existence lowers per capita wealth statistics, they're not "taking" that land from other people just by being alive - they have to work and trade for it like everybody else. That's capitalism.

Imagine what would happen if we dialed world population to extremes: 

Ok.

At one end, you would be the only person in the world. Theoretically, all the world's resources belong to you, making you very rich. But in practice, you will live an impoverished and short life, because you can't run a modern economy all by yourself to make use of those resources.

No, you're overstating the case. Most of those resources wouldn't belong to you; at least, they'd remain unavailable to you. This is an important distinction. You wouldn't even "own" any of those resources because you wouldn't have the manpower to actually dig them out of the ground or capture them. You would not be able to get any of the sunlight by building solar panels, or extract the coil and oil from the ground. The only resource you'd instantly "own" in a trivial sense would be the stuff that you could walk to and find some way to pick up, like the land in a certain radius around wherever you were.

When you imagine this scenario, I want to make sure you're not imagining a scenario where everybody in the world disappears, as that's not an analogous example. The actual world is filled with secondary goods that people built, not just premined coal and oil. The default one person world in your mind's eye should be Eliezer Yudkowsky transported to 10,000 B.C.E.

At the other end, the world is populated by 200 billion people and everyone is on the brink of starvation, surviving in urban slums on a diet of insects and algae.

This seems like an arbitrary number and I don't think you actually have any mechanical reason to believe this is the case. I would like to know what you believe are the bottlenecks on, for example, food production and city planning. Why wouldn't we be able to just scale up the solutions we already have? 

Somewhere in the middle, there is an optimum level of population that maximizes wealth per capita. This dynamic is a version of the Laffer curve, which tends to arise whenever two opposing factors are at play: For government tax revenue, it is the diminishing incentive to create wealth vs. the expanding fraction of wealth captured by the state. For aerodynamic resistance in rocketry, it is the decreasing air density vs. the increasing speed of the vehicle. For world's per capita income, it is the shrinking share of resources per person vs. the soaring economy of scale in production.

Again, this model doesn't account for the fact that new people can do things like build solar panels, farms, and oil drilling plants that actually "increase" the amount of "basic" resources that the rest of the world has access to. Innovations that increase the yield or make use of these resources better can also functionally be considered the same thing. 

In a further step, I believe there is overpopulation (i.e. oversupply of labor) at all levels, from unskilled labor all the way to the Einsteins of the world. I think the world's best scientists/engineers are mostly positional goods, similar to the world's best athletes - in a counterfactual world where they never existed, things would be chugging along at the same pace. This is why I'm not worried at all about the staggering number of university graduates and engineers China is producing each year - all this does is intensify internal competition.

You just state this is "what you believe" and don't explain why you believe it.  There are plenty of obscenely rich people in Silicon Valley today who would not win a positional game on business leadership in a million years but got lucky picking a correct idea. Larry and Sergei did not become billionaires by beating other entrepreneurs at a positional game - they became billionaires by creating a trillion-dollar search engine. And if your society mints 50% less entrepreneurs or engineers, you have 50% less opportunities for one of them to spend the trivial amount of resources coding and deploying a test version of Google. The faster the microchip/the microwave/the radio/etc. is developed the faster spinoff technologies can be developed and the more information the rest of the nation has on the search space for new tech. This is how "more people" turns into "proportionally more growth"; not only from the engineering and efficiency gains from regular engineers at Uber/AirBnB/Boeing but also the increased amount of rando startup founders who try their hand at making meme companies.

Perhaps you think scientific understanding is bottlenecked somehow by acceptance from a top cadre of "thought leaders", and I don't think I'd disagree in principle, but you also have to admit that it doesn't seem like the top minds of Ancient Greece don't seem like they would be able to handle building the atom bomb.

For this reason, I see the post-WW2 globalization of economy  as largely a redistribution scheme. I still think it was worthwhile, but only on moral grounds rather than economic self-interest of the powers that promoted globalization. Urbanization and rising living standards are among the most effective methods of hampering population growth, alleviating the problem at the source. Globalization has also been a promoter of peace, thus avoiding burning resources for nothing in wars.

I'm at a loss as to how this is supposed to follow. Do you think globalization causes poorer countries to increase in population and thus use up more global resources? But then this goes back to the fallacy that new people come with a share of world per capita income. The Chinese aren't using up land Americans owned before they started trading with them.

It's also just patently false in the short term - when the "post-WW2 globalization of the economy" ceased for Russia it made them poorer.

Here is a more relevant critique of your position: Notice however that the issue here is per capita income. The relevant metric for wartime capability is a nation's total productive capacity. Total productive capacity of a nation does increase with population and we can prove this with a simple thought experiment.

Point taken. Here is the problem with this argument: labor isn't the only input to productive capacity, and China's population is far passed the point where labor is the bottleneck. If China can't get enough oil, gas and other resources to fuel its industry and feed its people, it won't matter how many workers it has.

New babies are not born with a ticket that entitles them to a percent share of world GDP or land or iron ore. In practice they get a lot of support from their parents if they are to survive, but that's not really relevant for the purposes of this thought experiment because we can imagine them now at 18 years old and kicked out of the house with no money. Since on average these new people are still able to sell their labor on the market for more than it takes for them to survive, we can assume that they are contributing to China's economic size.

If you're going to use GDP as a proxy for power, well, I regret to tell you that China was by far the largest economy in the world in 1840 when they got their ass kicked by the British, and again in 1860. They lost to Japan in 1894 despite having 5 times the GDP.

No, most of those resources would remain unavailable to you. This is an important distinction. You wouldn't even "own" any of those resources because you wouldn't have the manpower to actually dig them out of the ground or capture them. You would not be able to get any of the sunlight by building solar panels, or extract the coil and oil from the ground. The only resource you'd instantly "own" in a trivial sense would be the stuff that you could walk to and find some way to pick up, like the land in a certain radius around wherever you were.

When you imagine this scenario, I want to make sure you're not imagining a scenario where everybody in the world disappears, as that's not an analogous example. The actual world is filled with secondary goods that people built, not just premined coal and oil. The default one person world in your mind's eye should be Eliezer Yudkowsky transported to 10,000 B.C.E.

That... is what I said with "you can't run a modern economy all by yourself"? Are you just arguing for the sake of arguing?

This seems like an arbitrary number and I don't think you actually have any mechanical reason to believe this is the case. I would like to know what you believe are the hard limits on, for example, food production and city planning. Why wouldn't we be able to just scale up the solutions we already have?

Correct, it is an arbitrary number. If you're not convinced that a world with 200 billion people and our current tech level is horrible to live in, take 2 trillion instead. The whole point is to dial the number to the extreme end of things, why bicker about this detail? The reason we can't just scale up current food production solutions, i.e. farmland, is because half the habitable land on Earth is already used for that purpose.

You just state this is "what you believe" and don't explain why you believe it.  There are plenty of obscenely rich people in Silicon Valley today who would not win a positional game on business leadership in a million years but got lucky picking a correct idea. And if your society mints 50% less entrepreneurs or engineers, you have 50% less opportunities for one of them to spend the trivial amount of resources coding and deploying a test version of Google. The faster the microchip/the microwave/the radio/etc. is developed the faster spinoff technologies can be developed and the more information the rest of the nation has on the search space for new tech.

I believe what I wrote because what you theorized has not come true in China or elsewhere. Shockingly little technological progress has happened in the last 20 years despite record levels of university graduates all around, this fact has merely been obfuscated by catch-up growth in China and other developing countries - a growth model that I predict can't be copied any more by e.g. India.

I'm at a loss as to how this is supposed to follow. Do you think globalization causes poorer countries to increase in population and thus use up more global resources? 

No, quite the opposite. Globalization reduces population by making poorer countries wealthier, which is good both directly and indirectly.

But then this goes back to the fallacy that new people come with a share of world per capita income. The Chinese aren't using up land Americans owned before they started trading with them.

They can compete for limited resources with their labor and thereby acquire a share of world per capita income.

Point taken. Here is the problem with this argument: labor isn't the only input to productive capacity, and China's population is far passed the point where labor is the bottleneck. If China can't get enough oil, gas and other resources to fuel its industry and feed its people, it won't matter how many workers it has.

As far as I can tell, we can settle this entire debate with one concrete question: of all the consumer goods China's/Your/My population(s) are buying, what percent of their value is due to the basic inputs like rented land and minerals, and what percent of their value is due to labor? Then apply your laffer curve and we will be able to calculate exaclty when new people "should" reduce per capita income statistics.

Correct, it is an arbitrary number. If you're not convinced that a world with 200 billion people and our current tech level is horrible to live in, take 2 trillion instead. The whole point is to dial the number to the extreme end of things, why bicker about this detail? The reason we can't just scale up current food production solutions, i.e. farmland, is because half the habitable land on Earth is already used for that purpose.

Well, partly because it sounded wrong and I like to bicker, and partly because the specific numbers are important. You're making a specific claim about 7 billion people being "too much"; I was wondering if you had any modeling behind this.

If you're going to use GDP as a proxy for power, well, I regret to tell you that China was by far the largest economy in the world in 1840 when they got their ass kicked by the British, and again in 1860. They lost to Japan in 1894 despite having 5 times the GDP.

I used the term "productive capacity" to more closely hint that that it's something more like "GDP per capita - subsistence income". Obviously if you're spending 90% of your economic and productive resources keeping your own population alive, or if your state doesn't have the political will/infrastructure to tax that income, then that GDP can't easily be reallocated towards wartime spending. In modern economies we have a lot more wiggleroom and so GDP itself more closely approximates "wartime production power".

Of course it's not the only thing. Governors can just fail to use the resources they have to protect their citizens. But spare change to buy guns, missiles, etc. is a resource.

That... is what I said with "you can't run a modern economy all by yourself"? Are you just arguing for the sake of arguing?

I'm doing this to point out that the "laffer curve" idea is flawed. Increases in population can also increase basic resources per capita by allowing society to scale up extraction and mint engineers that make efficiency and usage gains.

Doing multiple comments to keep track of different threads.

They can compete for limited resources with their labor and thereby acquire a share of world per capita income.

And again I'm confused. What's the problem here? In order to "acquire" that "share", China has to buy those resources from the United States, with something of equivalent or better value - mainly the fruits of their labor. They are not gifted that land from the United States as a trophy for being economically globalized, they enhanced their nation's productivity and now they are buying it with all of those secondary goods they're making, goods that we value as much more than the land itself. This is just a misunderstanding of how markets work.

To be clear: you think:

  1. Having more people to fight your war is a liability.
  2. At the same time, even though China is decreasing in population, "it's all downhill from here" for chinese military capability.
  1. Absolutely. More mouths to feed.
  2. As I already explained, in the long term the decreasing population is indeed a boon for China, but it takes several generations for this to take effect, and in the mid term they will be struggling with the burden of massive numbers of retirees just as Japan has struggled.

I'm thinking #1 cannot possibly be your actual view. Are you trolling? What do you think the "schelling" army+support staff size that can defeat any larger army is?

War isn't an RTS game, especially modern war. What do you expect the potentially 100 million strong Chinese army to do? Charge into Vietnam? Swim to the Strait of Malacca? China has always been the world's most populous nation yet that did nothing to change the fact that it basically lost every single war from 1840 - 1945. Or for that matter, look at how China was frequently raided and sometimes even conquered in its ancient history by numerically vastly inferior nomadic tribes.

I think everyone was surprised by how the rot of pervasive graft, corruption and theft ate into the Russian military. In 1941 the Soviet Army was ill-prepared and outdated, but corruption under Stalin was virtually non-existent. 

And whatever technical and strategic shortcomings became obvious in 1941, there was an existential threat and enough room to recover, if barely. The fall of Leningrad and Moscow was very close, as well, and with some better German planning (and not as harsh a winter) could have easily happened. Certainly if Germany decided to focus on beating Britain first and delay the Eastern operations by a few years, while getting oil and raw materials from all too willing Russia, the outcome might have looked very different. 

Basically, it was really hard to predict the current situation with any sort of confidence, and whatever experts happened to guess correctly are probably lucky more than anything.

In 1941 the Soviet Army was ill-prepared and outdated, but corruption under Stalin was virtually non-existent. 

This I think is a little misleading. Perhaps during the war itself the army shapened up, but by western liberal democratic standards Stalin and his general associates took tons and tons of money from their own people to spend on dachas and palaces. It was just not "corruption" because it wasn't "illegal".

As far as I know it is not true. I didn't know about any very spectacular palaces for Stalin or any his associates.

There was a corruption under Stalin. But as far as I know the main problem was false reports. Superiors demanded to fulfil (and to overfulfil) the plan, subordinates were afraid to confess that they cannot to do it and they was sending reports about a plan fulfilment.

They definitely got tons of money from the budget... I don't think it was stolen from military contracts, or whatever passed for military contracts in Stalin's economy.

the rot of pervasive graft, corruption and theft

This is intriguing, but I haven't seen any reporting on it. What are your sources? (That sounds combative on the Internet but is just me being curious.)

Your sources confirm that corruption is a problem, and it's plausible that corruption is a factor in how poorly the war has gone (which I note is the strongest claim, i.e. "plausible", in the Politico article), but your original claim, in the context of the OP you responded to, seemed to be that underestimation of corruption is [a huge part of? perhaps a majority of?] what caused everyone to be mistaken about Russian military power, and I definitely don't think these sources add up to that conclusion. 7 billion rubles of corruption in the military (Moscow Times article) is a drop in the bucket compared to a total budget of at least 2.5 trillion rubles, even if the corruption estimate is off by an order of magnitude.

The real corruption is and will always be maze-like behavior, not some guy embezzling a few billion dollars. Taxes we mortals can handle, mazes we cannot.

I think the common part in the current situation in Russian military and the situation before WWII is big problems with feedback. Before WWII in USSR you cannot tell about inabilty to fulfil a plan. And you had a very little chance to change superiors minds. Now in Russia you cannot criticize decisions of authorities efficiently. A criticism is a something which only western agents are doing.

But I think there is a distinction. I think a part of current problems in Russian military have emerged due to mass unbelief that a big war is possible. 

Basically, it was really hard to predict the current situation with any sort of confidence, and whatever experts happened to guess correctly are probably lucky more than anything.

I agree with this. My problem is not in the direction of my prediction, which I still think was correct ex ante, it's with the degree of confidence I had. I think predicting ~ 65% to 70% of the war going well for Russia with ~ 15% to 20% chance of a Winter War scenario would have been a better prediction in advance.

New to LessWrong?