This is very interesting! Thanks for taking the time to write it up.
I have a concern regarding the B-2 crash that happened over North Carolina. It seems to me that, if it is true that this lead to accidental nuclear detonations in most worlds, then we should expect to live in a world where the B-2 crash did lead to an accidental nuclear detonation but a nuclear holocaust was averted nevertheless.
My reasoning is as follows. In a fraction of worlds where nuclear detonation occurrs, we can expect that the American’s would recognize these detonations as accidental. Because the signs are there. For example, North Carolina does not seem like a prime target for a nuclear attack (though I’m not American, so maybe I’m mistaken about this). It would also seem odd that there were only two nukes—an attack from the Soviets would likely be bigger than that (reasoning along these lines is what caused Stanislav Petrov to recognize a false alarm in the Soviet missile detection system). The Americans might also remember that they had a B-2 flying over the over North Carolina. So in worlds where detonation occurs but the Americans pick up on these details, it is plausible that the Americans do not launch a retaliatory strike against the Soviets, and a nuclear war is averted so humanity can continue on without significant population loss.
Because of this, it seems like the observer selection effects should favor us living in a world where the B-2 crash lead to two accidental nuclear detonations in North Carolina without causing nuclear war. Because the suggestion seems to be that in most worlds this crash did lead to accidental detonation. But not all of these worlds experienced significant population loss. So depending on how we crunch the numbers, we may even expect most observers to exist in worlds where the B-2 crash lead to accidental nuclear detonation. But we do not live in one of these worlds. So arguably the observations we are making (of the B-2 crash not leading to detonation) are observations we are not likely to be making.
So here’s my question: do you think the fact that there wasn’t a nuclear detonation in North Carolina should lower our confidence in the Nuclear Fine Tuning Hypothesis?
I suspect that you might say no because the Nuclear Fine Tuning Hypothesis does not rely on the self-sampling assumption. As you mentioned, it only requires that we make descriptive rather than predictive claims. But it’s still not entirely clear to me how this distinction works. Because if we are explaining our current observations on account of them being probable on anthropic grounds, it seems unavoidable that we should predict that what we will observe will be probable on anthropic grounds too, simply because anthropic probability is on the table and the observations we make in the future are most likely going to be observations we are likely to make. Or at least that’s how it seems to me. I’d be interested to hear if/why you disagree with this.
I don’t maintain that the B-2 crash lead to nuclear detonation in most worlds, the examples are treated in the paper as straight coinflips but this is mostly to illustrate the general threat of accidental nuclear war. Zeroing in on this case I’d be surprised if it is more likely than not for this crash to lead to a nuclear detonation than not, while the safety mechanisms were inadequate for something as dangerous as a nuclear device they are still a relevant factor.
Throughout this whole project, this is a question that has continued to bug me. Would the US have had the ability to understand that it had nuked itself rather than blaming it on the Soviet Union? I honestly don’t know how to go about answering this question, my gut says no, and this is why there’s never been an accidental nuclear detonation on US soil. This is an assumption that carries through this paper, but I’d be interested to read a more nuanced take on it. It might also be that while the B-2 crash leading to a nuclear detonation wouldn’t have immediately led to nuclear war it would have increased tensions during the Cuban missile crisis(i.e. it might not immediately lead to nuclear war but it does decrease the odds). If the nuclear detonation(in the B-2 case) is even correlated with nuclear war then we should expect persons to disproportionately live in worlds where it doesn’t occur.
It is very weird that there has never been an accidental nuclear detonation on American soil, one accidental detonation would significantly change my assessment. As a result, I still think the B-2 crash is still evidence in favor of my hypothesis.
This is a promising counterargument, I think ultimately making it would require a complex assessment of US cold-war propaganda and deep dive into those in positions of power during different parts of the cold war.
I’m not exactly sure if my distinction between descriptive claims and predictive claims really works. My intuition is that descriptive claims are simply safer than predictive claims as they are less vulnerable to unseen factors. I will think about this more.
I think it is due to how nuclear war(in the past) would mean that I specifically wouldn’t exist. On the other hand if theirs a trillion people in the future I still would.
This is very interesting! Thanks for taking the time to write it up.
I have a concern regarding the B-2 crash that happened over North Carolina. It seems to me that, if it is true that this lead to accidental nuclear detonations in most worlds, then we should expect to live in a world where the B-2 crash did lead to an accidental nuclear detonation but a nuclear holocaust was averted nevertheless.
My reasoning is as follows. In a fraction of worlds where nuclear detonation occurrs, we can expect that the American’s would recognize these detonations as accidental. Because the signs are there. For example, North Carolina does not seem like a prime target for a nuclear attack (though I’m not American, so maybe I’m mistaken about this). It would also seem odd that there were only two nukes—an attack from the Soviets would likely be bigger than that (reasoning along these lines is what caused Stanislav Petrov to recognize a false alarm in the Soviet missile detection system). The Americans might also remember that they had a B-2 flying over the over North Carolina. So in worlds where detonation occurs but the Americans pick up on these details, it is plausible that the Americans do not launch a retaliatory strike against the Soviets, and a nuclear war is averted so humanity can continue on without significant population loss.
Because of this, it seems like the observer selection effects should favor us living in a world where the B-2 crash lead to two accidental nuclear detonations in North Carolina without causing nuclear war. Because the suggestion seems to be that in most worlds this crash did lead to accidental detonation. But not all of these worlds experienced significant population loss. So depending on how we crunch the numbers, we may even expect most observers to exist in worlds where the B-2 crash lead to accidental nuclear detonation. But we do not live in one of these worlds. So arguably the observations we are making (of the B-2 crash not leading to detonation) are observations we are not likely to be making.
So here’s my question: do you think the fact that there wasn’t a nuclear detonation in North Carolina should lower our confidence in the Nuclear Fine Tuning Hypothesis?
I suspect that you might say no because the Nuclear Fine Tuning Hypothesis does not rely on the self-sampling assumption. As you mentioned, it only requires that we make descriptive rather than predictive claims. But it’s still not entirely clear to me how this distinction works. Because if we are explaining our current observations on account of them being probable on anthropic grounds, it seems unavoidable that we should predict that what we will observe will be probable on anthropic grounds too, simply because anthropic probability is on the table and the observations we make in the future are most likely going to be observations we are likely to make. Or at least that’s how it seems to me. I’d be interested to hear if/why you disagree with this.
I don’t maintain that the B-2 crash lead to nuclear detonation in most worlds, the examples are treated in the paper as straight coinflips but this is mostly to illustrate the general threat of accidental nuclear war. Zeroing in on this case I’d be surprised if it is more likely than not for this crash to lead to a nuclear detonation than not, while the safety mechanisms were inadequate for something as dangerous as a nuclear device they are still a relevant factor.
Throughout this whole project, this is a question that has continued to bug me. Would the US have had the ability to understand that it had nuked itself rather than blaming it on the Soviet Union? I honestly don’t know how to go about answering this question, my gut says no, and this is why there’s never been an accidental nuclear detonation on US soil. This is an assumption that carries through this paper, but I’d be interested to read a more nuanced take on it. It might also be that while the B-2 crash leading to a nuclear detonation wouldn’t have immediately led to nuclear war it would have increased tensions during the Cuban missile crisis(i.e. it might not immediately lead to nuclear war but it does decrease the odds). If the nuclear detonation(in the B-2 case) is even correlated with nuclear war then we should expect persons to disproportionately live in worlds where it doesn’t occur.
It is very weird that there has never been an accidental nuclear detonation on American soil, one accidental detonation would significantly change my assessment. As a result, I still think the B-2 crash is still evidence in favor of my hypothesis.
This is a promising counterargument, I think ultimately making it would require a complex assessment of US cold-war propaganda and deep dive into those in positions of power during different parts of the cold war.
I’m not exactly sure if my distinction between descriptive claims and predictive claims really works. My intuition is that descriptive claims are simply safer than predictive claims as they are less vulnerable to unseen factors. I will think about this more.
I think it is due to how nuclear war(in the past) would mean that I specifically wouldn’t exist. On the other hand if theirs a trillion people in the future I still would.