Hacker News Reader

Did you read the entire article? He makes a point to discuss expectation values as the metric to judge the seriousness of addressing a nuclear event, not the statistical probability value

> Perry wishes more people were familiar with the concept of “expected value.” That is a statistical way of understanding events of very large magnitude that have a low probability. The large magnitude event could be something good, like winning a lottery ticket. Or it could be something bad, like a nuclear bomb exploding. Because the odds of winning the lottery are so low, the rational thing is to save your money and not buy the ticket. As for a nuclear explosion, by Perry’s lights, the consequences are so grave that the rational thing would be for people in the United States and everywhere to be in a state of peak alarm about their vulnerability, and for political debate to be dominated by discussion of how to reduce the risk.

Expected value breaks down with what I like to think of as the Pascal's wager fallacy: if you consider the consequences effectively infinite, any positive probability means you should think about nothing else.

It took a long time to get to the article's actual suggestions (eliminate all ICBMS?); until that, this was just fear mongering.

I'm not sure that I entirely agree with this assessment.

One of the very interesting things I got out of reading Command And Control by Eric Schlosser, was the number of very severe accidents that have happened in the US in the handling of nuclear weapons. While probably somewhat overstated, to me it basically makes a case that the US is lucky it hasn't nuked itself by accident (and some of the stories told in that book are incredible, I enjoyed that book).

Also, some of the stories make the case about how abysmal the care and prudence was around handling these weapons, such as the reluctance for years to even put codes on the weapons, or when they were forced to, setting every code to 0000. While I do believe the attitude may have changed, I don't know that it's prudent to put unwavering faith into the people handling these arms, based on what's told in this book.

While I don't want to say you suffer from this, I've seen what I believe to be a miss-allocation of risk in some of the engineering roles I've worked at. What I mean by this, is that because an event hasn't happened, or hasn't happened recently, that we believe the risk has changed. I've literally been part of a discussion on why do we need backups, because we've never restored from one.

If I understand correctly, this was the sort of philosophy that led to the challenger explosion, we launched the rocket before and it didn't explode, so why would it happen this time. To me, this is a fallacy of risk, previous events or recent history don't change the risk.

So while you're right, the risk that a nuclear accident happens, compared to a tornado or flooding happening is much different, doesn't mean that a nuclear accident won't happen. I also highly suspect, that if a nuclear armed country accidentally nukes itself, it would grip the world in a much different way than any tornado or flood. And may change your well-being in very dramatic way's, especially if one doesn't realize it was even an accident to begin with.

edit: minor grammatical mistakes

Also recommended reading is The Dead Hand by David E Hoffman.

The details about the zeroes:

http://arstechnica.com/tech-policy/2013/12/launch-code-for-u...

"Launch code for US nukes was 00000000 for 20 years"

http://foreignpolicy.com/2014/01/21/air-force-swears-our-nuk...

So it was not actually eight zeros (or four) but apparently 6 zeroes (000000) and effectively the "key under the doormat" (in the safe with the key in the same base as the rocket, the opposite of what was claimed then). The procedure was more complex than just keying in zeroes, but everything was surely trained on the rehearsals.

Now, imagine somebody just and only once forgets to apply the "it's just a rehearsal" command.

What's the "acceptable" probability some city being successfully targeted, even only and just once, especially knowing that it can trigger the retaliation?

The logic of the nuclear war is: "if the rockets come to you, you have to send yours to them, because otherwise they won!" -- they think it's better that the civilization gets wiped out than to appear to be a "chicken." Even a kind of semi-automated response surely exists.

>The logic of the nuclear war is: "if the rockets come to you, you have to send yours to them, because otherwise they won!" -- they think it's better that the civilization gets wiped out than to appear to be a "chicken." Even a kind of semi-automated response surely exists.

Isn't the idea behind Mutually Assured Destruction not some ego battle, but a deterrent? If nuking another country definitely results in one's own country being nuked, that provides a clear incentive towards self-preservation and not nuking that country in the first place.

It's actually very interesting. I don't remember if it was from Command and Control, or somewhere else, that I once read a story along the lines of.

When the missile launch crews, routinely were tested. And something along the lines of a set of codes was used for a test launch. And a different set of codes was used for a real launch. On one of these tests, the real launch codes were used by mistake. But due to the crew asking for confirmation several times, the person eventually noticed their mistake and cancelled the order. But if I understand/remember correctly, strictly speaking these crews were given a valid launch command.

While I didn't find my source for that particular example, I do believe it does happen where training and real have gotten mixed up. The most famous of which, is I think when the operator used a training tape in the early warning system, which caused the US to think they were under attack.

https://www.damninteresting.com/the-apocalypses-that-might-h...

> The logic of the nuclear war is: "if the rockets come to you, you have to send yours to them, because otherwise they won!" -- they think it's better that the civilization gets wiped out than to appear to be a "chicken." Even a kind of semi-automated response surely exists.

The semi-automated response is a good thing, it makes your commitment to retaliate even more certain.

This is because the enemy is typically expected to be aware of your capability, rational, and selfish. If they know that you won't fire the rockets back in response, they CAN win, and they might use their weapons. If they won't win, they won't fire the rockets in the first place.

Now, to be precise, the important part of the scheme is only that the enemy thinks you will fire the rocket. After the rockets are launched towards you, it doesn't matter anymore, and yes, you could (and should) nobly self-sacrifice your nation so that humanity survives. But any suggestion or expectation that this will happen makes it more likely that the enemy will call your bluff and destroy you.

Honestly, with this logic in consideration, it makes me wonder if some of the rumors and leaks about how close we have been to launching counterattacks are fabricated. Though given the risk of espionage, it's probably safer to simply not bluff.

All this falls apart if the enemy is either unaware of or disbelieving in your retaliation, so they might fire their missiles expecting you to not fire back. Or if they're irrational, failing to understand that they will not win. Or if they're not selfish (perhaps the scariest of the possibilities): Perhaps because they are suicidal, which could be considered a sub-class of irrational. Or because they think you will not retaliate against them, instead retaliating against someone else: For example, if through subterfuge the US could fire Pakistani missiles at North Korea, not being dissuaded by North Korea's retaliation against Pakistan because the destruction would be on a different continent, or for example the military or political leadership initiating the attack are in another country, or better yet, on another planet) and not personally harmed by their own country being destroyed.

I completely agree that the world would be a safer place if these weapons did not exist. Unfortunately, they do, and we're neither able to modify physics so that they don't, nor to unanimously agree that no use them, so we're stuck with this sub-optimal equilibrium.

>The logic of the nuclear war is: "if the rockets come to you, you have to send yours to them, because otherwise they won!"

That's not at all true. The logic of nuclear war is that your enemy must believe with a high degree of certainty that you will launch if attacked. Whether you actually do or not is beside the point.

I keep reading that the launch codes were set to zeroes in news articles but every technical document I read says the code was only used in Permissive Action Link devices. As I understand it, these PALs were mechanisms that prevented the nuclear weapon from arming so even if the code was accidentally put into the weapon it wouldn't actually detonate until the ICBM launch sequence was initiated (by a separate process) and the warhead detonation mechanism fired. They are a safety feature that was added on to nuclear weapons starting in the late 50's, not some sort of detonation or launch control system.

That said, the PALs weren't fitted to the entire US arsenal until the late 80's so lots of nukes had mechanical keys for decades. Hopefully they took the rehearsal part more seriously than the security of the arsenal...

Two serious fallacies here: 1) The risk posed by nuclear war is correlated across nearly every individual, so the odds ratio between risks posed to you as an individual and nuclear war is of no use to the practice of risk management in such policy matters. 2) You are comparing the study of things entirely outside our control (volcanos), and potential doom which is entirely of our own making, towards the end of maintaining control by holding the world under the constant threat of annihilation.

Are you sure the risk from tornadoes is higher? It looks like the odds of sustaining an injury or being killed by a tornado are somewhat under 0.1% per year for residents of Tornado Alley, and you can greatly reduce that for yourself with some simple precautions. I would guess that the risk of a nuclear holocaust is 0.1% per year or more, and there is nothing you can do to reduce the risk of it happening to you. (There are things you can do to increase your odds of survival and improve your likely position afterwards, but you will be severely impacted no matter what.)

> I would guess that the risk of a nuclear holocaust is 0.1% per year or more

> I would guess

Based on what? Unless it's based on something similar to where you got the 0.1% risk for a tornado it's no more accurate a guess than if you had said 0.01% or 1% or 80%.

Based on my understanding of the situation, and in particular based on several incidents over the past six decades where it was narrowly averted in situations where a few people having a bad day could have made it happen.

I'm not making any claim to accuracy. That's why I used the word "guess."

Note that the comment I replied to implies that the risk of a nuclear holocaust is "much lower" than 0.1% per year. Why aren't you hassling them about that claim too? Unlike me, they just stated it as if it were a plain fact.

He's a Bayesian, and his priors are predicated on the belief that the lack of apparent intelligent life in the universe is attributable to self-annihilation.

In all seriousness, calculating a probability of death from nuclear calamity isn't going to be possible without a great number of caveats and assumptions. It isn't like a meteor strike, because we at least have some record of that stretching back a long time. It's also not like meteor strikes because I believe the disaster generating process (people with buttons wired to faulty / hackable control systems) to be a giant mess of scary.

He didn't say '0.1% risk' he said it's 0.1% or more, with no upper bound. By giving a wide range he wasn't implying any sort of accuracy at all.

Based on several past incidents I would say we had a 50/50 shot of a nuclear holocaust over the last 70 years. That's around 1% per year. However, most of the incidents seem to be fairly old so we can probably drop that to 0.1% per year unless you live in India or Pakistan.

EX: "Solar Sunrise" is not that old and the US was seriously considering using nukes as NORAD was having major issues. IMO, large scale nuclear war was still unlikely, but the odds where probably well over 1% that year.

PS: Remember, WWI started from a chain of events that also seemed unlikely. But, a lot of stuff happens in any given year making such chains more common than you would think.

Think of it this way:

You are in a locked room with 9 idiots each holding a granade. What's your life expectancy?

Even worse, with even more particle physics, gene engineering, nanotechnology, artificial intelligence etc. you'll be soon locked up in a room with 7 billion idiots each holding a granade, a bazooka and a flame-thrower.

On one hand, you've been locked in the room with a slowly growing group of idiots for decades and nobody has started throwing grenades, at least not since John threw a couple of them when he got excited after being the first one in the room. On the other hand, Ivan keeps making strange faces at his grenade, Ali is playing "guess where I hid the grenade" and keeps getting into fistfights with Sanjay, Kim just barged in with a new grenade and keeps shouting at everyone, and John suddenly seems a lot more enthusiastic about blowing shit up real good.

> A nuclear catastrophe, statistically, poses a much lower risk to my well being than tornado season.

The probability may be lower. But the consequences are way, way, way, way, way bigger.(I am assuming nuclear catastrophe means nuclear war.)

From a personal point of view. A heart attack or cancer it is much bigger risk. You die and the world ends for you. But from the point of view of humanity. A nuclear war it is a much higher risk, since it would jeopardize our future.

the historical occurrences of nuclear device detonation have nothing whatsoever to do with the chances they will be used going forward.

i find his paranoia political. we should be a lot more worried about india and pakistan having nuclear weapons than a president who uses twitter.

Dismissive comments like these miss Perry's concerns entirely. We've been on this sleepwalk towards increasing nuclear risk long before Trump. There are many existing risks which are still inadequately addressed. Loose material being acquired by terrorists, North Korea, a Russia more openly using nuclear weapons as a threat (installing nuclear capable missiles in Kaliningrad as we speak!), the previous $1 Trillion US nuclear modernization program which while addresses many safety concerns introduces new tactical capabilities for weapons packages making them more usable.

He speaks out for concern on all of the above. The article states he hopes to meet with Trump & Mattis to work with him and give his advice.

my comment was to previous poster who used historical detonations as somehow predictive statistically. it isn't.

as to whether we should be concerned, where was this article five months ago? it's thinly veiled political commentary.

Perry has been speaking about this for many years.

http://www.wjperryproject.org/

If you want to disagree with him, then give an argument. However if you want to do a quick smear and dodge then keep repeating that it's "thinly veiled political commentary" over and over again.

Five months ago, there wasn't an increased vocalized threat of risk towards, for example, the Iran nuclear deal, or vocalized suggestions of nuclear arms proliferation increasing. There is new information now, and that needs to be factored into level of concern. Of course it's political, but that doesn't excuse new information from having an impact on risk assessment.

> where was this article five months ago?

Would three months ago count, or does it need to be exactly five? http://www.nytimes.com/2016/09/30/opinion/why-its-safe-to-sc...