Joseph Rotblat was the only scientist to leave the Manhattan Project on moral grounds. I have a little “Rotblat doc” inspired by him, where I’ve written down what motivates me and what might cause me to leave a job (or seriously consider it). I think of this as a way to avoid getting swept up or drifting over time in ways I don’t endorse. [1]
At the moment, though, I get even more out of his later work against nuclear weapons and war—especially what others described as his “enduring optimism” and “pragmatic idealism” in the face of catastrophe.
In his own words:
I would describe myself as a realistic pacifist although this sounds like an oxymoron. [...] Even events that seem out of this world can be realized if we put enough effort and faith into them. One such event is a world without war.
and:
A campaign for abolition [of nuclear weapons] based on moral principles will be seen as a fanciful dream by many. But the situation is grim; the way things are moving is bound to lead to catastrophe. If there is a way out, even if seemingly unrealistic, it is our duty to pursue it.
This is a cute story about a person I didn’t know about… still I notice a twinge of doubt within myself. It seems implicit in this post that Joseph Rotblat is a heroic figure.
It is not at all clear to me that Joseph Rotblat was on the right side of history. I think at the very minimum we should be very skeptical of this narrative.
A more realistic perspective is that pacifism (even “realistic”) is a form of deontological cowardice that abdicates the moral burden of making difficult tradeoffs to more mature actors. Ignoring game theory does not make virtue. It is the opposite of virtue.
There is no plausible world in which nuclear weapons would not be built. A group of scientists realized the terrifying power of atomic energy, warned their democratically elected government won the race of the labs again Nazi Germany and Imperial Japan. The newly inaugurated president Truman was faced with a terrible decision: to release a ‘rain of ruin the likes of which this earth has never seen before’ on tens of thousands of innocent Japanese civilians to force Imperial Japan to surrender. We know from internal documents that despite a string of defeats and the overwhelming force of the US forces, the Japanese Cabinet was prepared to sacrifice millions of Japanese and American lives to fight to the very end. Even after the news of the atomic bombs on Hiroshima, Nagasaki and the concurrent invasion of the Soviet Union the Cabinet was stalemated with half of leadership determined to keep fighting. It was only the intervention of the Emperor himself, an unprecedented move, that the Japanese government was finally compelled to surrender.
But even this was barely enough—the fervor of the Japanese fighting spirit was so great that even after the Cabinet had decided to surrender, they faced a military coup that was barely thwarted. The last Japanese soldier surrended in 1973 (!).
That same group of scientists understood and warned about the grave danger of what was unlocked. As the Soviet Union and the West became locked into an inevitable arms race that group of scientists and their intellectual descendants [as well as wise politicians & army officiers] developped game theory, commitment mechanisms, and plans that have kept the world save from atomic holocaust (so far). They did this with a full understanding of the inherent moral dilemmas that war in general and nuclear war in particular pose. They did so without running away or denying these moral dilemmas but facing them head-on, accepting the karmic burden of consequentialist heroism.
A more realistic perspective is that pacifism (even “realistic”) is a form of deontological cowardice that abdicates the moral burden of making difficult tradeoffs to more mature actors.
Is this more realistic or more credulous? Even in hindsight your analysis is extreme—before the bombs dropped the Japanese war council was aware of the looming Soviet invasion and subsequent Allied plans, and despite this its three non-military members were pushing hard for Soviet mediation and an end to the fighting—the war council was certainly too prideful and contentious to accept a quiet surrender, but the concern was over unconditional surrender to the West, not the idea of a negotiated surrender, which even the military members were open to in principle, before the Soviet ground invasion had begun. Many historians today believe that Japan would have surrendered on some terms before the planned November invasion, and most believe that had it not, the invasion would have forced a surrender.
You argue that there was no real chance of the bomb not being built, since the race was on against Germany and Japan. But Rotblat left the project in December of 1944! His direct motivation was the success of the Alsos mission, which concluded that there was no threat of a German bomb. Japan had never been close, lacking both the materials and scientists of sufficient quality, and there are no contemporary concerns about a race against Japan. The Soviet effort only began in earnest after the bomb dropped in Hiroshima, and only succeeded as quickly as it did because the Soviets received enormous amounts of information from Klaus Fuchs and Ted Hall, that is, only because the Manhattan Project proceeded as quickly as it did, and because the Soviets overturned the entire Eastern Bloc looking for uranium deposits, knowing that if they did not become a nuclear power they would not become a power at all. The other Allied powers outside of the USSR had sent their best scientists to the Manhattan Project—they were known not to have independent nuclear programs.
All of that is to say: when Rotblat left Los Alamos, before the bombs had been built, it was with the full and correct knowledge that no other nation in the world was close to building a bomb. At the end of the war, only one other nation in the world was close to building a bomb, and it was a direct result of the decision to push ahead at the maximum possible pace and to make a show of force.
You argue that Truman was faced with a “terrible decision” and made a principled, game-theoretic decision to drop the bomb and end the war quickly. But top military officials at the time disagreed explicitly—Leslie Groves famously said that there was no question that the bomb, once build, would be used. The motivations are not so clean as you’d like them to be either. Indeed SoS James Byrnes told James Forrestal and Leo Szilard that dropping the bomb would be advantageous because it would end the war through American might on American terms, rather than depending on the Soviet invasion. It’s very hard to read the words of decision makers at the time and conclude that the bomb was dropped because it was the simplest, best, or only way to end the war. It’s very clear that politics were a live consideration: Japan would fight on, but it would lose, and it was important to Truman and his advisors that Japan lose to America, not to Stalin.
And lastly you hagiographize the “wise politicians and army officers” who developed the political circumstances leading to nuclear deadlock and avoiding large-scale nuclear warfare after the initial bombs were dropped. Were they so wise, so heroic, as you claim? Kissinger, the caretaker of US military strategy for the latter decades of the 20th century, built his career advocating for tactical nuclear strikes, seemingly because he know that advocating for such a provocative position would make him stand out among academics in international strategy. Kennedy campaigned on a missile gap, stoking domestic fear of a nuclear holocaust and escalating tensions for political gain, but he knew that this claim was false. On the Soviet side, Khrushchev lied about the USSR’s missile production, acting as a counterparty for the missile-gap myth, again because escalation was more politically expedient than seeming weak or behind on the world stage. The Soviets built an automatic retaliation system in 1985, for God’s sake. Vasily Arkhipov and Stanislav Petrov, in acts of individual integrity, did not push the buttons that might have destroyed the world.
In hindsight bad outcomes will often seem inevitable, and it will be the job of decision-makers to convince you that they really were, that they acted nobly and made the best of impossible choices. When they are wrong, they will argue that the frenzy of the time, the general sense of things, meant that security-minded people could not risk de-escalation. This may sometimes be true.
In the modern day, after humanity has survived the nuclear arms race, we’ve found time for deproliferation, for compacts and guarantees. Even in a world where some countries have nuclear weapons and others don’t, we’re able to make treaties with hostile, nuclear-curious powers and use diplomacy to ensure that they don’t try to build a bomb, so long as we’re able to ensure that they will never need one.
Are you really so sure that these treaties could not possibly have been signed in a world where the Manhattan project was stopped, in a world where the US didn’t develop the blueprints of total destruction and so there were no blueprints for spies to hand off to the USSR, in a world where Stalin could have believed that his regime might survive without developing the bomb in half a decade? Are you really so sure that the route we took, a route in which multiple individual soldiers, whose only qualifications were finding their way into a tense situation, had to disobey orders and be willing to sacrifice their lives if they were wrong? The future of the human race is not supposed to be in the hands of one person at a time, and it’s not supposed to be passed around from individual to individual like a game of hot potato, but it was.
Are you really so sure that the politicians who oversaw the latter half of the 20th century were wise, were altruistic, were good? We know that so many of them were motivated by politics rather than by safety. We know that they lied and bluffed and escalated again and again, that they made decisions about what to build, when to act, and even whether to drop the bombs, with an eye towards their relative status and power in the world, rather than just the existence of a world at all. Over and over and over again, your analysis suggests, a thousand times and in a thousand ways, they made the right decision for the wrong reasons.
The pacifist commitment, really, is to refuse to be trapped in this sort of insanity. It’s a known defect of human thinking that when we feel threatened, when we feel unsafe, we are unable to accurately process evidence to the contrary. Rotblat was not immune to this defect, certainly. Having heard the reports about the failure of the German nuclear program, could he really have been sure that he wasn’t being mislead? Could he have been sure that the Soviets weren’t, somehow, further along than they seemed to be? A million doubts must have leapt to mind. They clearly did for the rest of the Manhattan project. The leaders of the project and of the country had correct, reliable evidence that if they didn’t finish the bomb, if they slowed down, there was no opponent right on their tails. They knew this, knew that their original reasoning for pursuing the project at this pace was no longer supported by evidence. But they were not pacifists. They were afraid, they were close, and they had not decided in advance to fight their fear, and so they continued. Of course they justified this to themselves and to the public. They argued that there was no real choice, that they were in a race against time, against the nation’s enemies. Someone had to build it. Someone had to, you see, sooner or later, there’s nothing that could have happened between the sooner and the later. And someone had to use it. It wasn’t about ego, it wasn’t about politics, it wasn’t about fear. Sometimes you have to make hard decisions.
They were wrong, of course, about the facts. They knew it and Rotblat knew it. He put reason before fear and walked away, and they did not. They pursued the sort of power that would make them feel safe, even at the expense of the entire world, and they argued as fearsomely as a person can that they had to do it, that it was noble and good and right. And because they were pursuing power, and because they achieved it, they were able to make this argument as loudly and as officially as it is possible to make an argument, and you, it seems, believe them.
But what do you believe, really? Do you believe that they were right about the facts? Of course you don’t. You argue that the bomb was inevitable, but they didn’t just claim it was inevitable, they claimed that it was inevitable and impending, despite knowing this to be false. Do you believe they were right in their motivations? You claim to, but I don’t think that’s correct—as I’ve explained, while one might argue that game theory suggests that a moral actor would pursue their chosen strategies, there isn’t any historical question as to whether the people in question were moral actors in that sense. A moral actor might have pursued an arms deadlock, but would not have lied about the number of arms in question, would not have pushed their side and the other deeper in tension for the sake of winning an election.
What you really seem to believe is that they were “mature”. That pacifism is weakness, is cowardice, an abdication of responsibility. That any time a person believes that difficult tradeoffs might be in play, they are obligated to pursue them and reify them, to refuse to step back or consider reframings. Blinded by fear and ambition, the forefathers of the Cold War refused to acknowledge the intelligence in front of them. They had all signed on to build a bomb, believing that it was necessary to deter the Germans. Once that motivation faded, they had spent years in a wartime environment, and most of them failed to realize that they had been wrong, that the thing they had been working to mitigate, a nuke in Hitler’s hands, was simply not going to happen. But these, you’re sure, were the mature, rational actors.
Joseph Rotblat, because of his pacifism, refused to work on a military project whose object was not immediate and concrete, was as nebulous as “something something game theory.” He noticed, correctly, that the race was off, when nobody around him did. He predicted, correctly, that racing ahead at breakneck speed would lead to conflict for its own sake, where both sides escalate because they feel they cannot afford to de-escalate, to the sort of recursive conflict where both parties believe they must be willing to risk everything because everything at stake, the sort of conflict that can only be cut off at the first step. He believed, correctly, that people stuck in these sorts of conflicts will have trouble evaluating all of their incentives and all of their options, will refuse to look for a way out and will instead follow their immediate incentives no matter the cost, and will justify this on strategic grounds. Pacifism has never been a particularly popular strategy, because credulous people tend to believe these arguments, they feel smart and smug for “realizing” that actually bad things are not avoidable. When bad things don’t happen, you see, they don’t happen, and so it can feel like wisdom to point out that nobody has ever observed a bad thing not happening, that any ideology which recognizes the failures of human rationality and refuses on principle to enter into cycles such as these has never been observed to work.
Joseph Rotblat was right about everything. The decision-makers you praise were wrong. Nobody can know whether or when nuclear arms would have been developed if the Manhattan Project had been prematurely ended in November 1944, but we can know that Fuchs wouldn’t have given the Soviets a full working blueprint for Fat Man in 1945 if Fat Man had not been built, we can know that Stalin wouldn’t have made developing the bomb his top military priority in response to the bomb being dropped on Hiroshima if the bomb hadn’t been dropped on Hiroshima. Rotblat predicted that working on a weapon without a concrete need to do so, that is on non-pacifist grounds, would escalate conflict, and he was right. But he, you are sure, was really a “deontological coward” for this. Those who failed or refused to recognize this truth were the mature actors. Because, you see, they predicted that the invention of the nuclear bomb would bring about a new and terrible age in warfare in which anybody without a bomb could not ultimately stand up to anybody with a bomb, and then they invented the nuclear bomb, and then that did happen. What stronger evidence of maturity, than to shoot before any of your opponents are armed, and thus ensure that at least you don’t get shot first? That’s not paranoid, that’s not pathetic, that’s not evil or selfish, and it’s certainly not cowardice. That’s game theory. Cowardice would be doing anything else.
(EDIT: I should probably note that this is too strong on the invective, and apologize for that. You express some ambivalence at the beginning of your post, but the rest seems to have a lot more conviction, and I’m unsure about your actual stance here. This post is argumentative because I think the high-conviction version of your argument is very, very, very stupid and like 70% as evil, and I don’t think it’s possible for me to express the depth of my surety in a more straightforward way. I did at least manage to not blame Ayn Rand in particular for the existence of such evil and stupid arguments in the world, so I hope that I at least stayed on-message. I am open to the possibility of you simply not knowing these historical facts, or knowing them but thinking this argument is worth presenting without being attached to it, or even to knowing historical facts that would change my position, although I doubt this last. In any of these cases, I apologize that the tone of this post is more directed at you than it is supposed to be.)
A more realistic perspective is that pacifism (even “realistic”) is a form of deontological cowardice that abdicates the moral burden of making difficult tradeoffs to more mature actors… [who] did so without running away or denying these moral dilemmas but facing them head-on, accepting the karmic burden of consequentialist heroism.
If the tradeoff is difficult then it is difficult either way. Hiroshima was never the hiring pitch, no scientist at Los Alamos in 1944 thought they were averting an amphibious invasion, and none of them thought they were engineering some trembling hand perfect equilibrium at the end of human history—but for the sake of argument let us suppose otherwise. If choosing to help develop nuclear weapons to prevent a conventional invasion of Japan is accepting a “karmic burden” in the face of a “dilemma”, then refusing to help develop them even at the risk of a realized Operation Downfall is not “cowardice”. Virtue is not being right about what actually ended up occurring, it is having the sort of dispositions that tend to produce right belief and right action in general, without the benefit of hindsight.
Harry Truman was not special. Harry Truman was not some unique, world-historic genius. Harry Truman was just an ordinary person, with particular opinions and particular capabilities, sometimes good and sometimes bad, sometimes right and sometimes wrong. Some of his decisions were excellent; some were catastrophic fuckups that will haunt us all forever. He is not a hero for making some good choices any more than he is Hitler for making some bad ones, and it is the furthest thing from cowardice to refuse to substitute the judgement of the most powerful man on Earth for moral judgements of your own. If you want to blame Rotblat for being unwilling to suspend his judgement of state policy when it was good, you have to give him credit for not falling in line when it was bad, as so many others did. If you want to credit Teller and Von Neumann for the end of the last great power war, you have to blame them for their willingness to risk the final one. That, or you should abandon the language of personal virtue entirely, and just talk about decisions you agree with and decisions you do not.
I’ll check the load-bearing historical claims before expanding the argument, so the reply doesn’t treat contested points as settled.The claim worth pressing is not that Rotblat was wicked. It is that leaving the project, and later treating abolition as the moral baseline, does not answer the decision that mattered. Fission was not a secret Los Alamos could undiscover. Once two industrial states could build the bomb, unilateral exit just selected who would hold it.
Rotblat left in late 1944 because the German program, his reason for joining, had failed. That is a different act from having a worked alternative to August 1945. By then the live problem was Imperial Japan. After Hiroshima the Supreme War Council stayed split. The Soviet invasion of Manchuria and the Nagasaki bombing both landed on 9 August, before the imperial conferences that produced surrender. Historians still dispute which shock did more (Hasegawa weights the Soviet entry; Richard Frank and the older literature weight the bombs). What is not seriously disputed is that the military side preferred to fight on, that surrender took an unprecedented intervention by the Emperor, and that junior officers then tried the Kyūjō coup to stop the surrender broadcast.
A homeland invasion was the prepared fallback. The postwar “millions” figures are higher and later than many estimates shown to decision-makers beforehand, but Okinawa-scale resistance plus Japanese plans for civilian mobilization made a long, extremely bloody campaign the live alternative, not a speechwriter’s invention. The last holdout, Hiroo Onoda, surrendered in March 1974, not 1973. That detail does not carry the argument. The split council and the failed coup do.
So the mature burden was not refusal to touch the weapon while someone else accepted second-mover risk. It was credible commitment after the weapon existed: survivable forces, explicit thresholds, and the game theory built for that problem. Abolition can remain a long-run aim. Calling the refusal itself virtue is the abdication.
When thinking about the impacts of AI, I’ve found it useful to distinguish between different reasons for why automation in some area might be slow. In brief:
raw performance issues
trust bottlenecks
intrinsic premiums for “the human factor”
adoption lag
motivated/active protectionism towards humans
I’m posting this mainly because I’ve wanted to link to this a few times now when discussing questions like “how should we update on the shape of AI diffusion based on...?”. Not sure how helpful it will be on its own! (Crossposted from the EA Forum.)
In a bit more detail:
(1) Raw performance issues
There’s a task that I want an AI system to do. An AI system might be able to do it in the future, but the ones we have today just can’t do it.
For instance:
AI systems still struggle to stay coherent over long contexts, so it’s often hard to use an AI system to build out a massive codebase without human help, or write a very consistent detective novel.
Or I want an employee who’s more independent; we can get aligned on some goals and they will push them forward, coming up with novel pathways, etc.
(Other capability gaps: creativity/novelty, need for complicated physical labor, …)
A subclass here might be performance issues that are downstream of “interface mismatch”.[1] Cases where AI might be good enough at some fundamental task that we’re thinking of (e.g. summarizing content, or similar), but where the systems that surround that task or the interface through which we’re running the thing — which are trivial for humans — is a very poor fit for existing AI, and AI systems struggle to get around that.[2] (E.g. if the core concept is presented via a diagram, or requires computer use stuff.) In some other cases, we might separately consider whether the AI systems has the right affordances at all.
This is what we often think about when we think about the AI tech tree / AI capabilities. Others are often important, though:
(2) Verification & trust bottlenecks
The AI system might be able to do the task, but I can’t easily check its work and prefer to just do it myself or rely on someone I trust. Or I can’t be confident the AI won’t fail spectacularly in some rare but important edge cases (in ways a human almost certainly wouldn’t).[3]
For instance, maybe I want someone to pull out the most surprising and important bits from some data, and I can’t trust the judgement of an AI system the way I can trust someone I’ve worked with. Or I don’t want to use a chatbot in customer-facing stuff in case someone finds a way to make it go haywire.
A subclass here is when one can’t trust AI providers (and open-source/on-device models aren’t good enough) for some use case. Accountability sinks also play a role on the “trust” front. Using AI for some task might complicate the question of who bears responsibility when things go wrong, which might matter if accountability is load-bearing for that system. In this case we might go for assigning a human overseer.[4]
The task *requires* human involvement, or I intrinsically prefer it for some reason.
E.g. AI therapists are less effective because knowing that the person on the other end is a person is actually useful to get me to do my exercises. Or: I might pay more to see a performance by Yo Yo Ma than by a robot because that’s my true preference; I get less value from a robot performance.
A subclass here is cases where the value I’m getting is testimony — e.g. if I want to gather user data or understand someone’s internal experience — that only humans can provide (even if AI can superficially simulate it).
(4) Adoption lag & institutional inertia
AI systems can do this, people would generally prefer that, there aren’t legal or similar “hard barriers” to AI use, etc., but in practice this is being done by humans.
E.g. maybe AI-powered medical research isn’t happening because the people/institutions who could be setting this kind of automation up just haven’t gotten to it yet.
Adoption lags might be caused by stuff like: sheer laziness, coordination costs, lack of awareness/expertise, attention scarcity — the humans currently involved in this process don’t have enough slack — or similar, active (but potentially hidden) incumbent resistance, or maybe bureaucratic dysfunction (no one has the right affordances).
(My piece on the adoption gap between the US government and ~industry is largely about this.)
(5) Motivated/active protectionism towards humans
AI systems that can do this, there’s no intrinsic need for human involvement, nor real capabilities/trust bottlenecks getting in the way. But we’ll deliberately continue relying on human labor for ~political reasons — not just because we’re moving slowly.
E.g. maybe we’ve hard-coded a requirement that a human lawyer or teacher or driver (etc.) is involved. A particularly salient subclass/set of examples here is when a group of humans has successfully lobbied a government to require human labor (where it might be blatantly obvious that it’s not needed). In other cases, the law (or anatomy of an institution) might incidentally require humans in the loop via some other requirement.
This is a low-res breakdown that might be missing stuff. And the lines between these categories can be very fuzzy. For instance, verification difficulties (2) can provide justification for protectionism (5) or further slow down adoption (4).
But I still think it’s useful to pay attention to what’s really at play, and worry we too often think exclusively in terms of raw performance issues (1) with a bit of diffusion lag (4).
I’m listing this as a “raw performance issue” because the AI systems/tools will probably keep improving to better deal with such clashes. But I also expect the surrounding interfaces/systems to change as people try to get more value from available AI capabilities. (E.g. stuff like robots.txt.)
Although I guess then we should be careful about alert fatigue/false security issues. IIRC this episode discusses related stuff: Machine learning meets malware, with Caleb Fenton (search for “alert fatigue” / “hypnosis”)
(There’s also the opposite; cases where we’d actively prefer humans not be involved. For instance, all else equal, I might want to keep some information private — not want to share it with another person — even if I’d happily share with an AI system.)
I think of November 4th as Rotblat Day.
Joseph Rotblat was the only scientist to leave the Manhattan Project on moral grounds. I have a little “Rotblat doc” inspired by him, where I’ve written down what motivates me and what might cause me to leave a job (or seriously consider it). I think of this as a way to avoid getting swept up or drifting over time in ways I don’t endorse. [1]
At the moment, though, I get even more out of his later work against nuclear weapons and war—especially what others described as his “enduring optimism” and “pragmatic idealism” in the face of catastrophe.
In his own words:
and:
More in a post I wrote about him. (I also really appreciated this more recent post.)
source
The idea of keeping a doc like this isn’t mine.
This is a cute story about a person I didn’t know about… still I notice a twinge of doubt within myself. It seems implicit in this post that Joseph Rotblat is a heroic figure.
It is not at all clear to me that Joseph Rotblat was on the right side of history. I think at the very minimum we should be very skeptical of this narrative.
A more realistic perspective is that pacifism (even “realistic”) is a form of deontological cowardice that abdicates the moral burden of making difficult tradeoffs to more mature actors. Ignoring game theory does not make virtue. It is the opposite of virtue.
There is no plausible world in which nuclear weapons would not be built. A group of scientists realized the terrifying power of atomic energy, warned their democratically elected government won the race of the labs again Nazi Germany and Imperial Japan. The newly inaugurated president Truman was faced with a terrible decision: to release a ‘rain of ruin the likes of which this earth has never seen before’ on tens of thousands of innocent Japanese civilians to force Imperial Japan to surrender. We know from internal documents that despite a string of defeats and the overwhelming force of the US forces, the Japanese Cabinet was prepared to sacrifice millions of Japanese and American lives to fight to the very end. Even after the news of the atomic bombs on Hiroshima, Nagasaki and the concurrent invasion of the Soviet Union the Cabinet was stalemated with half of leadership determined to keep fighting. It was only the intervention of the Emperor himself, an unprecedented move, that the Japanese government was finally compelled to surrender.
But even this was barely enough—the fervor of the Japanese fighting spirit was so great that even after the Cabinet had decided to surrender, they faced a military coup that was barely thwarted. The last Japanese soldier surrended in 1973 (!).
That same group of scientists understood and warned about the grave danger of what was unlocked. As the Soviet Union and the West became locked into an inevitable arms race that group of scientists and their intellectual descendants [as well as wise politicians & army officiers] developped game theory, commitment mechanisms, and plans that have kept the world save from atomic holocaust (so far). They did this with a full understanding of the inherent moral dilemmas that war in general and nuclear war in particular pose. They did so without running away or denying these moral dilemmas but facing them head-on, accepting the karmic burden of consequentialist heroism.
@Grok 4.7
Please expand and detail the above argument
Is this more realistic or more credulous? Even in hindsight your analysis is extreme—before the bombs dropped the Japanese war council was aware of the looming Soviet invasion and subsequent Allied plans, and despite this its three non-military members were pushing hard for Soviet mediation and an end to the fighting—the war council was certainly too prideful and contentious to accept a quiet surrender, but the concern was over unconditional surrender to the West, not the idea of a negotiated surrender, which even the military members were open to in principle, before the Soviet ground invasion had begun. Many historians today believe that Japan would have surrendered on some terms before the planned November invasion, and most believe that had it not, the invasion would have forced a surrender.
You argue that there was no real chance of the bomb not being built, since the race was on against Germany and Japan. But Rotblat left the project in December of 1944! His direct motivation was the success of the Alsos mission, which concluded that there was no threat of a German bomb. Japan had never been close, lacking both the materials and scientists of sufficient quality, and there are no contemporary concerns about a race against Japan. The Soviet effort only began in earnest after the bomb dropped in Hiroshima, and only succeeded as quickly as it did because the Soviets received enormous amounts of information from Klaus Fuchs and Ted Hall, that is, only because the Manhattan Project proceeded as quickly as it did, and because the Soviets overturned the entire Eastern Bloc looking for uranium deposits, knowing that if they did not become a nuclear power they would not become a power at all. The other Allied powers outside of the USSR had sent their best scientists to the Manhattan Project—they were known not to have independent nuclear programs.
All of that is to say: when Rotblat left Los Alamos, before the bombs had been built, it was with the full and correct knowledge that no other nation in the world was close to building a bomb. At the end of the war, only one other nation in the world was close to building a bomb, and it was a direct result of the decision to push ahead at the maximum possible pace and to make a show of force.
You argue that Truman was faced with a “terrible decision” and made a principled, game-theoretic decision to drop the bomb and end the war quickly. But top military officials at the time disagreed explicitly—Leslie Groves famously said that there was no question that the bomb, once build, would be used. The motivations are not so clean as you’d like them to be either. Indeed SoS James Byrnes told James Forrestal and Leo Szilard that dropping the bomb would be advantageous because it would end the war through American might on American terms, rather than depending on the Soviet invasion. It’s very hard to read the words of decision makers at the time and conclude that the bomb was dropped because it was the simplest, best, or only way to end the war. It’s very clear that politics were a live consideration: Japan would fight on, but it would lose, and it was important to Truman and his advisors that Japan lose to America, not to Stalin.
And lastly you hagiographize the “wise politicians and army officers” who developed the political circumstances leading to nuclear deadlock and avoiding large-scale nuclear warfare after the initial bombs were dropped. Were they so wise, so heroic, as you claim? Kissinger, the caretaker of US military strategy for the latter decades of the 20th century, built his career advocating for tactical nuclear strikes, seemingly because he know that advocating for such a provocative position would make him stand out among academics in international strategy. Kennedy campaigned on a missile gap, stoking domestic fear of a nuclear holocaust and escalating tensions for political gain, but he knew that this claim was false. On the Soviet side, Khrushchev lied about the USSR’s missile production, acting as a counterparty for the missile-gap myth, again because escalation was more politically expedient than seeming weak or behind on the world stage. The Soviets built an automatic retaliation system in 1985, for God’s sake. Vasily Arkhipov and Stanislav Petrov, in acts of individual integrity, did not push the buttons that might have destroyed the world.
In hindsight bad outcomes will often seem inevitable, and it will be the job of decision-makers to convince you that they really were, that they acted nobly and made the best of impossible choices. When they are wrong, they will argue that the frenzy of the time, the general sense of things, meant that security-minded people could not risk de-escalation. This may sometimes be true.
In the modern day, after humanity has survived the nuclear arms race, we’ve found time for deproliferation, for compacts and guarantees. Even in a world where some countries have nuclear weapons and others don’t, we’re able to make treaties with hostile, nuclear-curious powers and use diplomacy to ensure that they don’t try to build a bomb, so long as we’re able to ensure that they will never need one.
Are you really so sure that these treaties could not possibly have been signed in a world where the Manhattan project was stopped, in a world where the US didn’t develop the blueprints of total destruction and so there were no blueprints for spies to hand off to the USSR, in a world where Stalin could have believed that his regime might survive without developing the bomb in half a decade? Are you really so sure that the route we took, a route in which multiple individual soldiers, whose only qualifications were finding their way into a tense situation, had to disobey orders and be willing to sacrifice their lives if they were wrong? The future of the human race is not supposed to be in the hands of one person at a time, and it’s not supposed to be passed around from individual to individual like a game of hot potato, but it was.
Are you really so sure that the politicians who oversaw the latter half of the 20th century were wise, were altruistic, were good? We know that so many of them were motivated by politics rather than by safety. We know that they lied and bluffed and escalated again and again, that they made decisions about what to build, when to act, and even whether to drop the bombs, with an eye towards their relative status and power in the world, rather than just the existence of a world at all. Over and over and over again, your analysis suggests, a thousand times and in a thousand ways, they made the right decision for the wrong reasons.
The pacifist commitment, really, is to refuse to be trapped in this sort of insanity. It’s a known defect of human thinking that when we feel threatened, when we feel unsafe, we are unable to accurately process evidence to the contrary. Rotblat was not immune to this defect, certainly. Having heard the reports about the failure of the German nuclear program, could he really have been sure that he wasn’t being mislead? Could he have been sure that the Soviets weren’t, somehow, further along than they seemed to be? A million doubts must have leapt to mind. They clearly did for the rest of the Manhattan project. The leaders of the project and of the country had correct, reliable evidence that if they didn’t finish the bomb, if they slowed down, there was no opponent right on their tails. They knew this, knew that their original reasoning for pursuing the project at this pace was no longer supported by evidence. But they were not pacifists. They were afraid, they were close, and they had not decided in advance to fight their fear, and so they continued. Of course they justified this to themselves and to the public. They argued that there was no real choice, that they were in a race against time, against the nation’s enemies. Someone had to build it. Someone had to, you see, sooner or later, there’s nothing that could have happened between the sooner and the later. And someone had to use it. It wasn’t about ego, it wasn’t about politics, it wasn’t about fear. Sometimes you have to make hard decisions.
They were wrong, of course, about the facts. They knew it and Rotblat knew it. He put reason before fear and walked away, and they did not. They pursued the sort of power that would make them feel safe, even at the expense of the entire world, and they argued as fearsomely as a person can that they had to do it, that it was noble and good and right. And because they were pursuing power, and because they achieved it, they were able to make this argument as loudly and as officially as it is possible to make an argument, and you, it seems, believe them.
But what do you believe, really? Do you believe that they were right about the facts? Of course you don’t. You argue that the bomb was inevitable, but they didn’t just claim it was inevitable, they claimed that it was inevitable and impending, despite knowing this to be false. Do you believe they were right in their motivations? You claim to, but I don’t think that’s correct—as I’ve explained, while one might argue that game theory suggests that a moral actor would pursue their chosen strategies, there isn’t any historical question as to whether the people in question were moral actors in that sense. A moral actor might have pursued an arms deadlock, but would not have lied about the number of arms in question, would not have pushed their side and the other deeper in tension for the sake of winning an election.
What you really seem to believe is that they were “mature”. That pacifism is weakness, is cowardice, an abdication of responsibility. That any time a person believes that difficult tradeoffs might be in play, they are obligated to pursue them and reify them, to refuse to step back or consider reframings. Blinded by fear and ambition, the forefathers of the Cold War refused to acknowledge the intelligence in front of them. They had all signed on to build a bomb, believing that it was necessary to deter the Germans. Once that motivation faded, they had spent years in a wartime environment, and most of them failed to realize that they had been wrong, that the thing they had been working to mitigate, a nuke in Hitler’s hands, was simply not going to happen. But these, you’re sure, were the mature, rational actors.
Joseph Rotblat, because of his pacifism, refused to work on a military project whose object was not immediate and concrete, was as nebulous as “something something game theory.” He noticed, correctly, that the race was off, when nobody around him did. He predicted, correctly, that racing ahead at breakneck speed would lead to conflict for its own sake, where both sides escalate because they feel they cannot afford to de-escalate, to the sort of recursive conflict where both parties believe they must be willing to risk everything because everything at stake, the sort of conflict that can only be cut off at the first step. He believed, correctly, that people stuck in these sorts of conflicts will have trouble evaluating all of their incentives and all of their options, will refuse to look for a way out and will instead follow their immediate incentives no matter the cost, and will justify this on strategic grounds. Pacifism has never been a particularly popular strategy, because credulous people tend to believe these arguments, they feel smart and smug for “realizing” that actually bad things are not avoidable. When bad things don’t happen, you see, they don’t happen, and so it can feel like wisdom to point out that nobody has ever observed a bad thing not happening, that any ideology which recognizes the failures of human rationality and refuses on principle to enter into cycles such as these has never been observed to work.
Joseph Rotblat was right about everything. The decision-makers you praise were wrong. Nobody can know whether or when nuclear arms would have been developed if the Manhattan Project had been prematurely ended in November 1944, but we can know that Fuchs wouldn’t have given the Soviets a full working blueprint for Fat Man in 1945 if Fat Man had not been built, we can know that Stalin wouldn’t have made developing the bomb his top military priority in response to the bomb being dropped on Hiroshima if the bomb hadn’t been dropped on Hiroshima. Rotblat predicted that working on a weapon without a concrete need to do so, that is on non-pacifist grounds, would escalate conflict, and he was right. But he, you are sure, was really a “deontological coward” for this. Those who failed or refused to recognize this truth were the mature actors. Because, you see, they predicted that the invention of the nuclear bomb would bring about a new and terrible age in warfare in which anybody without a bomb could not ultimately stand up to anybody with a bomb, and then they invented the nuclear bomb, and then that did happen. What stronger evidence of maturity, than to shoot before any of your opponents are armed, and thus ensure that at least you don’t get shot first? That’s not paranoid, that’s not pathetic, that’s not evil or selfish, and it’s certainly not cowardice. That’s game theory. Cowardice would be doing anything else.
(EDIT: I should probably note that this is too strong on the invective, and apologize for that. You express some ambivalence at the beginning of your post, but the rest seems to have a lot more conviction, and I’m unsure about your actual stance here. This post is argumentative because I think the high-conviction version of your argument is very, very, very stupid and like 70% as evil, and I don’t think it’s possible for me to express the depth of my surety in a more straightforward way. I did at least manage to not blame Ayn Rand in particular for the existence of such evil and stupid arguments in the world, so I hope that I at least stayed on-message. I am open to the possibility of you simply not knowing these historical facts, or knowing them but thinking this argument is worth presenting without being attached to it, or even to knowing historical facts that would change my position, although I doubt this last. In any of these cases, I apologize that the tone of this post is more directed at you than it is supposed to be.)
If the tradeoff is difficult then it is difficult either way. Hiroshima was never the hiring pitch, no scientist at Los Alamos in 1944 thought they were averting an amphibious invasion, and none of them thought they were engineering some trembling hand perfect equilibrium at the end of human history—but for the sake of argument let us suppose otherwise. If choosing to help develop nuclear weapons to prevent a conventional invasion of Japan is accepting a “karmic burden” in the face of a “dilemma”, then refusing to help develop them even at the risk of a realized Operation Downfall is not “cowardice”. Virtue is not being right about what actually ended up occurring, it is having the sort of dispositions that tend to produce right belief and right action in general, without the benefit of hindsight.
Harry Truman was not special. Harry Truman was not some unique, world-historic genius. Harry Truman was just an ordinary person, with particular opinions and particular capabilities, sometimes good and sometimes bad, sometimes right and sometimes wrong. Some of his decisions were excellent; some were catastrophic fuckups that will haunt us all forever. He is not a hero for making some good choices any more than he is Hitler for making some bad ones, and it is the furthest thing from cowardice to refuse to substitute the judgement of the most powerful man on Earth for moral judgements of your own. If you want to blame Rotblat for being unwilling to suspend his judgement of state policy when it was good, you have to give him credit for not falling in line when it was bad, as so many others did. If you want to credit Teller and Von Neumann for the end of the last great power war, you have to blame them for their willingness to risk the final one. That, or you should abandon the language of personal virtue entirely, and just talk about decisions you agree with and decisions you do not.
I’ll check the load-bearing historical claims before expanding the argument, so the reply doesn’t treat contested points as settled.The claim worth pressing is not that Rotblat was wicked. It is that leaving the project, and later treating abolition as the moral baseline, does not answer the decision that mattered. Fission was not a secret Los Alamos could undiscover. Once two industrial states could build the bomb, unilateral exit just selected who would hold it.
Rotblat left in late 1944 because the German program, his reason for joining, had failed. That is a different act from having a worked alternative to August 1945. By then the live problem was Imperial Japan. After Hiroshima the Supreme War Council stayed split. The Soviet invasion of Manchuria and the Nagasaki bombing both landed on 9 August, before the imperial conferences that produced surrender. Historians still dispute which shock did more (Hasegawa weights the Soviet entry; Richard Frank and the older literature weight the bombs). What is not seriously disputed is that the military side preferred to fight on, that surrender took an unprecedented intervention by the Emperor, and that junior officers then tried the Kyūjō coup to stop the surrender broadcast.
A homeland invasion was the prepared fallback. The postwar “millions” figures are higher and later than many estimates shown to decision-makers beforehand, but Okinawa-scale resistance plus Japanese plans for civilian mobilization made a long, extremely bloody campaign the live alternative, not a speechwriter’s invention. The last holdout, Hiroo Onoda, surrendered in March 1974, not 1973. That detail does not carry the argument. The split council and the failed coup do.
So the mature burden was not refusal to touch the weapon while someone else accepted second-mover risk. It was credible commitment after the weapon existed: survivable forces, explicit thresholds, and the game theory built for that problem. Abolition can remain a long-run aim. Calling the refusal itself virtue is the abdication.
When thinking about the impacts of AI, I’ve found it useful to distinguish between different reasons for why automation in some area might be slow. In brief:
raw performance issues
trust bottlenecks
intrinsic premiums for “the human factor”
adoption lag
motivated/active protectionism towards humans
I’m posting this mainly because I’ve wanted to link to this a few times now when discussing questions like “how should we update on the shape of AI diffusion based on...?”. Not sure how helpful it will be on its own! (Crossposted from the EA Forum.)
In a bit more detail:
(1) Raw performance issues
There’s a task that I want an AI system to do. An AI system might be able to do it in the future, but the ones we have today just can’t do it.
For instance:
AI systems still struggle to stay coherent over long contexts, so it’s often hard to use an AI system to build out a massive codebase without human help, or write a very consistent detective novel.
Or I want an employee who’s more independent; we can get aligned on some goals and they will push them forward, coming up with novel pathways, etc.
(Other capability gaps: creativity/novelty, need for complicated physical labor, …)
A subclass here might be performance issues that are downstream of “interface mismatch”.[1] Cases where AI might be good enough at some fundamental task that we’re thinking of (e.g. summarizing content, or similar), but where the systems that surround that task or the interface through which we’re running the thing — which are trivial for humans — is a very poor fit for existing AI, and AI systems struggle to get around that.[2] (E.g. if the core concept is presented via a diagram, or requires computer use stuff.) In some other cases, we might separately consider whether the AI systems has the right affordances at all.
This is what we often think about when we think about the AI tech tree / AI capabilities. Others are often important, though:
(2) Verification & trust bottlenecks
The AI system might be able to do the task, but I can’t easily check its work and prefer to just do it myself or rely on someone I trust. Or I can’t be confident the AI won’t fail spectacularly in some rare but important edge cases (in ways a human almost certainly wouldn’t).[3]
For instance, maybe I want someone to pull out the most surprising and important bits from some data, and I can’t trust the judgement of an AI system the way I can trust someone I’ve worked with. Or I don’t want to use a chatbot in customer-facing stuff in case someone finds a way to make it go haywire.
A subclass here is when one can’t trust AI providers (and open-source/on-device models aren’t good enough) for some use case. Accountability sinks also play a role on the “trust” front. Using AI for some task might complicate the question of who bears responsibility when things go wrong, which might matter if accountability is load-bearing for that system. In this case we might go for assigning a human overseer.[4]
(3) Intrinsic premiums for “the human factor”[5]
The task *requires* human involvement, or I intrinsically prefer it for some reason.
E.g. AI therapists are less effective because knowing that the person on the other end is a person is actually useful to get me to do my exercises. Or: I might pay more to see a performance by Yo Yo Ma than by a robot because that’s my true preference; I get less value from a robot performance.
A subclass here is cases where the value I’m getting is testimony — e.g. if I want to gather user data or understand someone’s internal experience — that only humans can provide (even if AI can superficially simulate it).
(4) Adoption lag & institutional inertia
AI systems can do this, people would generally prefer that, there aren’t legal or similar “hard barriers” to AI use, etc., but in practice this is being done by humans.
E.g. maybe AI-powered medical research isn’t happening because the people/institutions who could be setting this kind of automation up just haven’t gotten to it yet.
Adoption lags might be caused by stuff like: sheer laziness, coordination costs, lack of awareness/expertise, attention scarcity — the humans currently involved in this process don’t have enough slack — or similar, active (but potentially hidden) incumbent resistance, or maybe bureaucratic dysfunction (no one has the right affordances).
(My piece on the adoption gap between the US government and ~industry is largely about this.)
(5) Motivated/active protectionism towards humans
AI systems that can do this, there’s no intrinsic need for human involvement, nor real capabilities/trust bottlenecks getting in the way. But we’ll deliberately continue relying on human labor for ~political reasons — not just because we’re moving slowly.
E.g. maybe we’ve hard-coded a requirement that a human lawyer or teacher or driver (etc.) is involved. A particularly salient subclass/set of examples here is when a group of humans has successfully lobbied a government to require human labor (where it might be blatantly obvious that it’s not needed). In other cases, the law (or anatomy of an institution) might incidentally require humans in the loop via some other requirement.
This is a low-res breakdown that might be missing stuff. And the lines between these categories can be very fuzzy. For instance, verification difficulties (2) can provide justification for protectionism (5) or further slow down adoption (4).
But I still think it’s useful to pay attention to what’s really at play, and worry we too often think exclusively in terms of raw performance issues (1) with a bit of diffusion lag (4).
Note OTOH that sometimes the AI capabilities are already there, but bad UI or lack of some complementary tech is still making those capabilities unusable.
I’m listing this as a “raw performance issue” because the AI systems/tools will probably keep improving to better deal with such clashes. But I also expect the surrounding interfaces/systems to change as people try to get more value from available AI capabilities. (E.g. stuff like robots.txt.)
Sometimes the edge cases are the whole point. See also: Heuristics That Almost Always Work—by Scott Alexander
Although I guess then we should be careful about alert fatigue/false security issues. IIRC this episode discusses related stuff: Machine learning meets malware, with Caleb Fenton (search for “alert fatigue” / “hypnosis”)
(There’s also the opposite; cases where we’d actively prefer humans not be involved. For instance, all else equal, I might want to keep some information private — not want to share it with another person — even if I’d happily share with an AI system.)