« What it's like to talk to your voice | Main | Why I like Facebook »
Robin Hanson crushes the Doomsday Argument
Robin writes:
It is interesting that doomsday argument proponents seem to challenge our usual way of doing inference, by preferring an extended state space where we explicitly model the idea that "I could have been you." However, if we try to do this in a physics-oriented way, avoiding describing states directly in abstract features of interest to humans but not the universe, we get seem to get the same chance of doom as if we hadn't extended states at all. Humanity may in fact face doom soon, and we have many reasons to be concerned about this. But I do not think the doomsday argument is one of them.
Here is the full argument. This piece is not new, but I believe most of you do not know it. Here is a previous MR post on the Doomsday Argument, also not supportive.
The bottom line: You still have to save for your retirement.
Posted by Tyler Cowen on August 28, 2007 at 01:19 PM in Philosophy | Permalink
Comments
My problem with the Doomsday Argument is that it misuses probability.
Given that you are number X in the number of humans that have lived, what is the probability that Y number of human beings (Y > X) will ever live?
Why, exactly the same as if I were any other number. The probability of me being a certain number is very very small, but the probability of me being SOME number is 1:1.
Put it another way: given that the bicycle you own has sequential serial number 243, what is the probability that the manufacturer will produce approximately 300 bicycles?
The information is worthless. Given worthless information, the likelihood of an expected outcome does not change. It's completely irrelevant what the serial number of your bicycle is. The idea that anyone remotely intelligent could possibly be convinced by this argument astounds me.
Posted by: Caliban Darklock at Aug 28, 2007 1:56:51 PM
I also should have mentioned the excellent papers by Nick Bostrom, see the left hand side of MR for the link to his home page.
Posted by: Tyler Cowen at Aug 28, 2007 2:26:07 PM
I should add that many others have offered similar critiques of the doomsday argument. Even if I'm mostly not persuaded by the argument, it has been a very valuable contribution that has forced us to be more clear about why the argument is wrong.
Posted by: Robin Hanson at Aug 28, 2007 3:47:27 PM
Hm, what about a retirement doomsday argument? How many days until enough net-taxpayers abandon countries
that keep jacking up taxes to pay for ever-increasing social pension obligations, that they seize
anyone's retirement savings they can find, regardless of previous promises?
Posted by: Person at Aug 28, 2007 3:51:25 PM
The Doomsday Argument requires societal mortality--speaking of the global population collectively as a society.
Planetary mortality we can observe in other planets and therefore expect for our own. However, we have no sample from which to guess patterns of societal mortality. Doom is therefore imbedded in the premises of the argument and can surprise no one when it arises in the conclusion. That said, the only meaning left for the argument is to say that the supposed doom is imminent. Such a result seems completely dependent on the supposed distribution of finite societies' lifespans. We have observed only one society and cannot tell whether it will be finite or infinite. So again, I see no basis for determining a distribution of societal lifespans.
John Adams said to Thomas Jefferson regarding their shared belief in the immortality of the human soul, "If we are wrong, we shall never know it." Indeed, regardless of how little some might see themselves as oriented toward faith, Paul's words apply as well to an elite atheist as to an ignorant Christian: "We walk by faith, not by sight".
Posted by: Matthew Petersen at Aug 28, 2007 4:09:22 PM
Put it another way: given that the bicycle you own has sequential serial number 243, what is the probability that the manufacturer will produce approximately 300 bicycles?
Lower than if it were number 299, which is in turn lower than if it were number 301. I'm not sure if this validates the Doomsday Argument.
Posted by: Nick Tarleton at Aug 28, 2007 7:03:28 PM
Robin's arguments are okay but I think the basic premise of the Doomsday Argument holds, it's just a matter of *quality* of assumptions, the assumptions are not wrong overall.
Anyway, bashing the DA is kind of lame, as is supporting it, what would be really helpful is if Tyler would state his expected median humanity life expectancy.
Posted by: Paul N at Aug 28, 2007 8:59:13 PM
Every time that I've seen the doomsday argument advanced, it has been immediately followed by, "so what causes the crash?" Which is then followed by the preachers preaching that if we don't change feature X of our societal behavior, then humanity will be wiped out by the year twenty-five twenty-five. Such claims of course fail to observe that the doomsday scenario computations are independent of the proposed action.
Singularity or bust!
Posted by: Nathan Zook at Aug 29, 2007 8:53:24 AM
I think that the DA can call on one "real" datapoint. This is the fact that SETI is so far a complete failure.
If the typical path of inteligent life was always onward and upward, as Robin believes, wouldn't we have already met our new posthuman (oops I mean postorganic) overlords?
Posted by: RobbL at Aug 29, 2007 9:41:28 AM
This is peculiar because the "we are living in a simulation" argument of Bostrom and Hanson is really just a version of the doomsday argument.
Posted by: Alex Tabarrok at Aug 29, 2007 4:10:30 PM
Alex, the simulation and doomsday arguments really are pretty different in their premises, even if they have related conclusions.
Posted by: Robin Hanson at Aug 29, 2007 4:40:54 PM
The doomsday argument is only about births, not deaths. Supposing effective anti-aging therapies are developed and both the birth and death rates decline to very low levels? Would not this fit the probability math of the doomsday argument?
Posted by: Kurt9 at Aug 30, 2007 1:21:46 AM
One addendum I haven't seen commented on is that the Doomsday argument requires that human inter-generational lifetimes be a constant (say 25 years).
If the inter-generational lifespan were to increase, say (through the magic of The Singularity or otherwise) to 100, 1000 years, or even indefinitely, then it quite is possible to be in the last few generations of (post)humans without having "Doom Soon".
Posted by: Alistair Morley at Aug 31, 2007 10:56:20 PM
I'd also like to add that Robin Hanson's treatment is similar to a lot of Bayesian analysis which clarifies some of the problems in the DA. I don't think there's anything really difficult with the problems he identifies, but here's my statisticians take on them:
1) It's naughty to infer global distributions of X without stronger grounds (and how do we know the overall human population is temporally exponential anyway? What's wrong with a long slow decline or stability as in Western Europe?)
2) Even if you _know_ global distribution of X, then all knowledge of your local position (x1) does is enable you to update your estimate of doomsday (x2); the median human life between (x1) and (Xmax). If Xmax is large compared to x1 (and many "reasonable" values are), then Doom is unlikely anytime soon.
3) Confusion in sampling estimation between to doomsday and to doomsday. Remember; the doomsday argument assumes a sample from an ordering of all human lives under a hypothetical exponential with a known lamda and upper bound. Less well noticed is the DA assumption of uniformity for the incidence of Doomsday is in the same human population NOT Doomsday incidence is uniform in time. This might be best phrased as "Everyone has the same chance of been the last human born" NOT "Doomsday is uniformerly distributed between X0, Xmax"
If we do have an upper bound (Xmax in mind), then our (updated) estimate of doomsday should be that it falls between my life (x1) and the max possible human lives (Xmax). Note that on an exponential curve this "median life" will be much closer (on the time axis) to Xmax than X1, our current position. NOT half-way on the time axis. This is exactly the reverse of the normal inference suggested by this argument!
4) The assumption about mean duration of human lifespan, mentionned above.
Posted by: Alistair Morley at Aug 31, 2007 11:36:03 PM
I don't buy it. Humanity will not bring about its own demise through natural causes (i.e. over-population). The only way that humanity can be wiped out (before the sun goes nova that is) is through nuclear war.
my world political forum
Posted by: roger at Sep 3, 2007 3:57:50 AM
China Tours and Yangtze Cruises A professional China travel agency offering private & group package tours, tailor-made vacations for families, couples and individual travelers, Yangtze cruises, hotel & ticket booking services.
China Travel and Tours One-stop China travel service delivering China package tours, discount Yangtze River cruises and best rate hotels.
Tibet Train Travel A reliable tour operator specializing in Tibet tour packages, adventures, hotel and train ticket booking.
China Maps A free web-based source for China travel and tour map users
Voyage Chine, Tour de Chine, Circuit Chine - Un agence spécialiste voyage en Chine
Great Wall of China A website for the Great Wall of China.
Posted by: China tour at Nov 27, 2007 2:18:45 AM
Hi! i hope to find a regular, good interior design室內設計. services company, I will let my new home beautiful, warm.I have already agreed the move搬家公司 as long as the decorations finished, we can immediately move into new homes.
Posted by: Tony at Dec 28, 2007 9:36:37 PM
Knight Online Gold
Knight Gold
Knight Noah
Knight Online Noah
Posted by: aion kina at Mar 19, 2009 10:45:35 PM
