The book Black Swan counters fallacies like this reasoning.
1. Since his example (the fall of the Berlin Wall) is a one time event without any historical data and can only happen once, it's glorified guessing.
2. He ignores that a bell curve is specifically for results that have very predictable ranges. In Black Swan, Taleb calls this "Mediocristan" because the range of height of adult human males holds roughly to a predictable range aka we don't run into men who are 6 inches or 10 feet tall.
3. By taking this once in a lifetime event and applying a bell curve to it - as if the existence Berlin Wall was a gaussian event - he's trying to make guessing look like science.
4. After declaring it "proven", he goes on to compare it to another one time event.. then end of human life.
If he was talking about a regular event with historical data, this may be useful analysis but as is.. oh, he's selling a book. Got it.
On the contrary, the article never mentioned bell curve, gaussian, or normal distribution. You're projecting your own assumptions onto it. The reasoning may be suspect, but it makes no assumptions about distributions and you've misunderstood it entirely.
> "Because it can be applied to completely arbitrary distributions provided they have a known finite mean and variance, the inequality generally gives a poor bound compared to what might be deduced if more aspects are known about the distribution involved."
In the situations cited in the article, we are dealing with both an unknown mean and an unknown variance and therefore an unknowable distribution. So while Chebyshev is "weaker" than a Gaussian distribution, it's still a distribution.
Sorry, I was not saying that the article applied Chebyshev's inequality. It's not related at all. But it's an example of a result where you do not have to assume that your distribution is a member of a particular family of distributions. The result being applied here is really really general: it says that 95% of the probability mass occurs before the point at which 95% of the probability mass has occurred by. This is tautological. Like saying that 50% of people are below the median - it's true without making any assumptions about people (except that there are finite numbers of them).
Wow. That's truly pinpointing it! "Between 5k and 8m years from now." If you're willing to accept "between 100 and 10b years from now", it's not-quantitatively-but-qualitatively equivalent to any living human but even better odds, I'd put them at 99.9%.
Something seems fishy about applying this math here. The human population is vastly bigger today than at any point in history. We don't know how big it will be in future. As an individual human trying to estimate your chances of being alive today this would seem to throw off the statistics. I'm much more likely to be alive today as one of 7 billion than 200,000 years ago as one of probably much less than a million. On the other hand it we go on to colonize the galaxy for a billion years I'm much less likely to be alive now than at the peak of our galactic empire.
That's a good point - the distribution over time should affect the answer.
I've heard this rule of thumb applied before to the same problem in a different way - rather than assuming you're somewhere in the middle of the time period humans exist for, assume you're somewhere in the middle of the list of all humans in order of birth.
(This version probably makes a much more pessimistic prediction than the article does?)
The argument here is basically the "Doomsday argument" seen elsewhere except applied to to time instead of population. If you do this argument but say that we're likely in the middle 95% of the total human population, then your complaint will go away. However, that doesn't give you time estimates, so it's maybe not as intuitive a result as the one in this article.
My bogometer started ringing immediately. The article even calls out what the problem is, but offers no resolution or explanation of how it doesn't kill the whole hypothesis. The problem is that the figures you get are profoundly affected by how far you are from the start. How does this not erase all of the presumed information content?
This is nonsense. The probabilistic theory suggested might apply well to the examples they gave of Broadway plays and world leaders in power, which both have natural limits on their timespans, but it's not at all obvious that applies well to guessing the length of human civilisation. Crucially that is not well informed by its length so far; neither of the threats they quoted earlier (nuclear annihilation and catastrophic global warming) have been a potential factor for more than 0.1% of the 200,000 year figure they're using as our lifespan as a species so far.
And they ask in what way we are different to Neanderthals or Homo erectus in terms of longevity; well neither of them ever came even slightly close to wiping themselves out with nuclear weapons, or developing space travel to allow them to survive global catastrophes. Comparing their longevity might be useful to predict how long we would survive from tens of thousands of years ago when our ancestors lived in caves in Africa, but it's utterly different to trying to predict the longevity of the civilisation we have now.
In Sheri S Tepper's wonderful novel Sideshow (the final part of the Arbai trilogy) a planet of (increasingly rare and nativist) baseline humans is set up to answer the question: "What is the destiny of man?"
The answer comes back: "To become something that is not man."
Extinction itself isn't really anything to be scared of. The real question is whether we're replaced by something better, and maybe what those superior beings would think of us after we're gone.
(That last question is answered in one of Stephen Baxter's books with: "an extremely successful animal that enlarged its ecological niche")
>Extinction itself isn't really anything to be scared of.
It is to those getting extinct. And whether they're replaced by "something better" or not doesn't matter to that.
Either it's you (your species) that is evolving, or it matters very much.
In a hypothetical scenario, I supposed few people would rush to go voluntarily extinct to save the life of some more evolved (smarter, stronger, more beautiful, etc) alien species out there.
No more than ordinary death is. Some extinction scenarios have us as the (possibly biological) progenitors of our successors. You'd still want what was best for your children. The ability to interbreed with baseline humans would be at best a secondary concern.
Anyway, I highly recommend the Arbai trilogy. Grass, in particular, is a masterpiece.
As pointed out by others in this thread, this is basically the doomsday argument all over again.
It's a deep subject, and I can't pretend I know much about it, but one thing that I'd like to point out is that this kind of reasoning is based of the so-called "self-sampling assumption"[1], and that this concept depends of a choice of reference class.
Even if Richard Gott is right, what he means when he's talking about "humans" going extinct is "people that I can identify myself with" going extinct. In other words : he can only make assessments on the existence of people like him.
Humans are not perfect. Far from it. If anything, the existence of exceptionally smart people like Einstein or Von Neuman prove us that it's possible to imagine a world where everybody is at least as smart as those two. Arguably, that hypothetical future world may very well be outside of Gott's reference class. Such a world could result from the birth of a new Homo species and the end of Sapiens. It could mean that machines would have replaced us. It could mean lots of things, not all of them being necessarily dreadful.
My point being : the doomsday argument is not exactly a prediction about the demise of mankind, but rather one about a dramatic change of it. In a way, it's more apocalyptic in the original sense of the world : not the end of times, but a profound change, a new era or whatever. It's a prediction about the end of our reference class. Or in the Kurzweilian sense, the Singularity.
My favorite way to end civilization would be a gamma ray burst. It travels at the speed of light, so there's no way to know it's even approaching. And it'd probably irradiate the whole solar system, so a Mars colony wouldn't save us.
Hmm... What are some other tempting apocalyptic scenarios? Asteroids are so 1998.
One thing that's bugged me is being born so close to the invention of nuclear weapons. If we're a random sample in all of human history, then why were we born so close to the birth of nukes? If humans are going to be around for another few million years, it's a minuscule chance that we should be born right now instead of much later.
You might think there have been a hundred billion humans born, so that logic doesn't really work: it's a one-in-a-hundred-billion chance that we were born here anyway. But the population is spread out in a J curve over the course of eons. It's not evenly distributed. https://en.wikipedia.org/wiki/Doomsday_argument
It's fun to waste time thinking about such things.
We are not a random sample in all human history. We are, by definition, the sample of population in human history alive at this moment.
Us being here contemplating it says nothing of the probability of being born now vs some other time. Other than that its probability is now 1 because it has already happened.
I think his argument is wrong and the fallacy comes when making the hypothesis the is a "50 percent chance that it was somewhere in the middle portion of the wall's timeline". It seems like a relatively reasonable assumption in that context but it is actually very arbitrary.
Imagine applying his reasoning to any building that has just been completed. Say we visit it the following day, make the 50% hypothesis and then conclude the building has a 50% chance of falling within 8 hours to 3 days. Obviously the conclusion is wrong as vastly more than 50% of buildings last longer than 3 days. We see in this case that postulating that there is a 50% chance that we visited the building somewhere in the middle portion of the building's timeline was completely unreasonable - we know from experience that buildings tend to last years.
So basically his estimation of when humans go extinct derives from a guess that may or may not be correct and in any case seems very arbitrary.
Another way of looking at it: instead of imagining you're at a random time in humanity, imagine you're a random human. The number of humans on earth has been increasing exponentially, so if you pick one at random they are most likely going to be from the last two centuries. This skews the results by making it more likely that the prediction is being made at this particular point in time (or a bit later if the population continues growing), which means that it is a non-uniform sample.
If all the visitors to that building applied this logic, then 50% of them would correctly believe that they were among the middle 50% of visitors to the building. In that sense it is correct reasoning without making any assumptions.
Alternative conclusion (not sure if the reasoning is correct) is that people will reach longevity relatively soon so new people won't be born anymore.
Anyway, it's really fascinating if you think about this. We're living in possibly the most unique and interesting century in the 200k year history of mankind - and even in the 4B years history of life - what are the chances of that? We're near the peak rate of technological progress.
What we consider normal, constant technological and societal progress is unique in the big picture, it can't continue forever and will stabilize in a few thousands years. After that, there will be long periods of stability and little change. The year 738,842 won't be dramatically different from 1,738,938.
His reasoning applied to the length of Broadway plays, or the length of political terms, is valid because such events have been going on for a long time, there have been multiple occurrences and they have a naturally limited length. Many have already ended, so it can be considered that we are in a "steady state" and statistical considerations about the events can be made. In the case of singular events however like the extinction of the Human species there is no such assumption. However when he uses the "95% rule" he implicitly assumes a normal distribution of events. That render his conclusion invalid due to use of an invalid starting hypothesis. In the case of the Berlin wall, he was just lucky.
Firstly, you should be considering the effect and not the object. The Berlin Wall was just a wall. It was geo-politics that bought down the wall. That's what was being predicted.
These changes tend to happen on human timelines. So arguing that an 8-year old geo-political setup may change in the next 24 years would make sense. That's basically saying in the next generation or two. You can throw the statistics out.
The effects that could wipe out humanity is a relatively specific list. And they definitely don't occur in human timeframes. You can work probabilities off that list. You'll come up with a completely different answer.
>We haven't put anyone on the moon since 1972, for instance. As a percent of gross domestic product, NASA's budget has been slowly petering out since the late 1960s. It's quite possible that “there will only be a brief window of opportunity for space travel during which we will have the capability to establish colonies,” Gott wrote in 1993. “If we let that opportunity pass without taking advantage of it we will be doomed to remain on Earth where we will eventually go extinct.”
So only the USA and only NASA can save all of humanity?
"But as Gott points out, our Neanderthal ancestors were around for only 300,000 years, while Homo erectus survived for about 1.6 million. They were smarter than the animals around them, but from a longevity standpoint they were completely unremarkable. Why should we be any different? Why should we be special?"
Circular argument. Competition with us was a factor in neanderthals extinction.
1. Since his example (the fall of the Berlin Wall) is a one time event without any historical data and can only happen once, it's glorified guessing.
2. He ignores that a bell curve is specifically for results that have very predictable ranges. In Black Swan, Taleb calls this "Mediocristan" because the range of height of adult human males holds roughly to a predictable range aka we don't run into men who are 6 inches or 10 feet tall.
3. By taking this once in a lifetime event and applying a bell curve to it - as if the existence Berlin Wall was a gaussian event - he's trying to make guessing look like science.
4. After declaring it "proven", he goes on to compare it to another one time event.. then end of human life.
If he was talking about a regular event with historical data, this may be useful analysis but as is.. oh, he's selling a book. Got it.