Dear Max Tegmark,
I read your risk assessment with an abstract mathematical scenario model with great interest, but was also very worrying, because your result does not bode well and gives no prospect of a possible happy ending.
However, I see a number of errors in the approach or better it does not reflect our current world situation. There's just a bunch of undisclosed facts that put everything in a different light. In your statement, dear Max Tegmark, these developments in the world have not been taken into account. Your model turns out to be like a model for the Cuban Missile Crisis. So it's stuck with a look back at the sixties. It's like Hans Rosling's book "Factfulness": we always have to learn that the world is different today than it was in the past.
I believe that, on the one hand, we a r e already in the midst of a war between East and West; only the war is conducted differently today. And furthermore, both sides have a "new general" sitting at the war map tables today: I consider the development of artificial intelligence to be so advanced that it is definitely used for tactical consultation and strategic planning, and if that is not the case yet, it will shortly be involved in war planning and decisions.
Of course, Russia has similar support, only they are in a far weaker position as the US leads by miles in AI development. Alpha-Zero, the super chess-playing, self-learning system, like Alpha-Star, is ideal for transferring to "similar problems". And a constantly changing war situation poses similar problems. Here, too, it's basically about the best use of matter, space and time; and about tactics and strategies. The self-learning system is fed with data from space surveillance, the news from Ukraine and Russia and learns extremely quickly how wars are waged and won by absorbing all previous wars and battles in history, recorded on the Internet and in literature, and uses millions of battle simulations against itself for achieving highest levels to the point that his tactics and strategy are largely invincible, at least bringing Russia to the edge of defeat. At the same time, all measures and recommendations must be dosed in such a way that a psychological system component keeps Putin's potential decision on a nuclear strike as low as possible.
I think that's how warfare works today. And it is real, no SF.
And we saw the first results recently, with everyone wondering how Ukraine managed to launch such a successful counteroffensive.
Of course, a nuclear escalation by Russia cannot be ruled out 100%, but the probability is well below your 1:6 model.
If that were to happen, the possible consequences would be Armageddon, perhaps like Harlan Ellison's 1968 Hugo-winning story "I have no mouth and I must scream" (https://en.wikipedia.org/wiki/I_Have_No_Mouth, _and_I_Must_Scream) described. The status of today's AI leads not only to the best game players, whether chess, go or video games, the best medical scientists or chemical, biological inventors and future predectors, but also to the best war advisors, which the Pentagon has likely just integrated into the General Staff. Only with AI the Ukraine problem can and will be solved. And we will see that soon. At least in 2023 you will remember my words.
Rudolph Angeli, Hamburg
Many people have asked me what I think the odds are of an imminent major US-Russia nuclear war. My current estimate is about the same as losing in Russian roulette: one in six. The goal of this post is to explain how I arrived at this estimate. Please forgive its cold and analytic nature despite the emotionally charged topic; I'm trying not to be biased by hopes, fears or wishful thinking.
My estimate is 30% x 80% x 70% ~ 1/6, as illustrated in the figure and explained below. The horizontal axis roughly corresponds to levels of escalation, while the vertical axis corresponds to how favorable outcomes are to the two sides.
To estimate the odds of pulling a spade out of a deck of cards, it's important to know how many suits there are. To estimate the odds that the current unstable situation ends up in the "KABOOM" outcome (a major US-Russia nuclear war that might cause nuclear winter and kill most people on Earth), it's similarly important to know what other reasonably stable outcomes it's competing against. The shorthand labels I've given these outcomes (grey boxes) should't be taken too literally: "Kosovo" & "Vietnam" refer to scenarios where one side wins outright (breakaway succeeds & Goliath is expunged, respectively). "Libya", "Korea" & "Finland" refer to intermediate outcomes involving simmering war, frozen war and full peace, respectively. I'm not showing the "Cuba" outcome (invasion averted by negotiated agreement) that was on the table in December 2021, since it's now off the table, as are resumed EU-Russia gas exports via the Nordstream pipelines.
Escalation dynamics
The grey ellipses represent relatively short-lived situations. We are currently in a vicious circle in the form of a self-perpetuating escalation spiral: since "Kosovo" is deemed unacceptable by Ukraine and the West while "Vietnam" is deemed unacceptable by Russia, both sides double down and escalate further whenever they fear losing. Such escalation has been both quantitative (more weapons, more mobilization) and qualitative (e.g., novel sanctions, heavier weapons, longer-range weapons, attacks inside Russia, scaled-up attacks on civilian infrastructure, shelling of a nuclear power plant, assassinations, sabotage of gas pipelines and Europe's longest bridge, annexations, and escalatory rhetoric about nuclear use). My assessment is that Russia, whose GDP is similar to Italy's, can no longer compete with the West in terms of quantitative escalation, and that Putin understands that his only chance to avoid the "Vietnam" outcome is to escalate qualitatively, with nuclear weapons use being his last resort. Last spring, I predicted that once loss of occupied territory loomed, he would annex what he controlled and start talking about nuclear defense of Russia's new borders – and here we are.
Breaking the vicious circle
I view it as highly unlikely (<10%) that Putin would accept "Vietnam" without first going nuclear, because it would almost certainly result in him being overthrown and jailed or killed. On the other hand, I also view it as highly unlikely (<10%) that the West would accept a "Kosovo" scenario where Russia is granted a peace deal where it keeps everything it's annexed, because if the powers that be in the West were that appeasement-minded, they would presumable have opted for a "Cuba" scenario in 2021 by acquiescing to Russia's demand that Ukraine never join NATO. This means that with high (>80%) probability, the current vicious cycle of escalation will end only with de-escalation into one of the intermediate outcomes ("Libya"/"Korea"/"Finland") or with lower-case "kaboom" (Russian nuclear use in Ukraine).
Estimates of the "kaboom" probability have recently ranged from 5% to 9% in the Metaculus prediction community. My current estimate is a few times higher (30%, e.g. a 2-to-1 chance that the cycle will end with de-escalation rather than escalation), because de-escalation currently seems so disfavored: there appears to be a widespread assumption in the West, shared by Ukrainian leaders, that Ukraine is winning and that Putin will grudgingly accept "Vietnam". Moreover, there is a near-consensus in mainstream Western media and policy circles against peace negotiations, exemplified by e.g. the hostile response to Elon Musk's recent suggestion of a peace deal.
Post-nuclear escalation
The probability that "kaboom" (nuclear use in Ukraine) leads to "KABOOM" (WW3) obviously depends on the Western response and subsequent escalation dynamics. My estimate is quite high (80%) that NATO's response will be forceful enough to include a non-nuclear military strike against Russia, because key NATO leaders and others have already made strongly worded statements to this effect. Options discussed have included sinking Russia's Black Sea fleet, which it would be difficult to imagine Russia not viewing as a declaration of war. My most likely (70%) scenario after that is Russian counterstrikes followed by rapid escalation via retaliatory actions from both sides, culminating in execution of the all-out nuclear war plans that both sides have spent decades preparing. My 70% estimate factors in that the long history of nuclear near misses has convinced me that both the US and Russia are much less competent in de-escalation than in escalation.
In the slightly less likely (30%) scenario that global freakout brings the US and Russia back from the brink, de-escalating toward the left side of the diagram, the outcome may be closer to "Kosovo" or "Vietnam" depending on who blinks first, i.e., on whether the de-escalation happens after "kaboom" or "Expansion".
WW3 impact
Many detailed estimates of nuclear war impact have been published in the academic literature. Xia et al (Nature Food, 3, 586–596, 2022) estimate that nuclear winter would kill about 99% of all Russians, Americans, Europeans and Chinese, with the most powerful post-war remaining economies being in South America, Southern Africa and Oceania. However, more work is needed to reduce uncertainties e.g. targeting scenarios, black carbon smoke production and lofting.
The only nuclear target map thus far declassified by the United States suggested that China would also be targeted even in a US-Russia war, to prevent it from emerging as the strongest post-war economy. My guess is that such a strategy is in force today as well, given the frosty state of Sino-US relations. Since China has much more large cities than either the US or Russia, this significantly increases my smoke production estimate.
I'd love to hear your thoughts both on this risk modeling framework and on the factor probabilities (30%, 80%, 70%) listed in the figure! I'll plan to update them regularly as the geopolitical shituation evolves.
De-escalation clarification
Many Twitter responses to this post have conflated nuclear de-escalation with capitulation or appeasement. Conversely, not all escalation has military value. For example, goading Putin to escalate with Moscow car bombing or viral video taunts is arguably against the national security interests of Ukraine and the West. If you're generally opposed to de-escalation, I'm curious as to which of the following escalations you don't want both sides to stop:
1) nuclear threats
2) atrocities
3) assassinations lacking military value
4) infrastructure attacks lacking military value (e.g. Nordstream sabotage)
5) shelling the Zaporizhzhya nuclear plant
6) misleading disparagement of de-escalation supporters as unpatriotic or appeasement-seeking