Hacker News Reader

Comment by user-the-name | Hacker News Reader

> I see no fallacies

"I am more likely to have been born in a high population country than a low population country"

This is incorrect. If there is one country with one million people, and a hundred countries with a hundred thousand people, you are overwhelmingly more likely to have been born in a low population country.

View on HN

Only if you aren't careful enough about what you mean by "low population". The argument is that you are probably born in a country with above the median population.

This will also be the case in your example. The median population is 100K. Assume there are tiny deviations around the median. Then 50 x 100K people live in countries below the median population. And 49 x 100K people, plus 1 million, live in countries above the median. (And 100K people live in the median population country itself.)

If you don't like the tiny deviations, then still, more people live in countries weakly above the median (everyone) than live in countries weakly below the median (everyone minus a million); and more people live in countries strictly above the median than strictly below.

But does "above the median population" actually tell you anything that is useful? I don't see that as being anything more than a truism.

The claim being made here is what I call the "Existential inference hypothesis": that we are somehow, as individuals, sampled from a large ensemble of consciousness across space and time. There is some difficulty in defining the conscious cross-section that governs this sampling process (could you be born a mouse? Maybe yes, but then you probably couldn't be asking those questions... Could you be born a rock? That makes even less sense; but all that differs among those cases is their atomic configuration). I think the sampling in this case might be a soft measure among linguistically capable entities.

What it does is allow some weak inference about the existential landscape, about everything that exists out there. I believe the author's refutal of 'inference from a single datapoint' is mostly wrong -- I believe indeed we're likely some kind of uniform sample, and single uniform samples don't reveal variance. But, there's not exactly a single variable here. We have a pretty large number of variables: population of the Earth, number of Neurons per individual, and so on. It's still not much though to give an accurate picture of the landscape.

I would indeed be wary of taking any hard conclusions from the data. If you use hard knowledge from biology, physics and computational disciplines, you can probably use our "sample" to refine estimates about life in our own universe.

Overall, the most I would conclude from this hypothesis is the following: "You and I probably (but not certainly) typical among numerical living beings in the entirety of existence.". The practical lesson I would take from that is to be thankful for what you have, and do the most of what you can for the world, because the basic donitions don't get much better than this.

We really can't draw any even weak inferences here. The only thing we know is that life that is much like ourselves is possible. We can not in any way say if it is common, or so exceedingly unlikely that we are the only example. The only data point we have is that we exist, but that data point confers no information, since it is a given. If we did not exist, we could not consider it.

So we know that one example of life exists, the Earth. The question we are trying to answer is, does a second example exist. For this, we have no information other than that it is possible.

No, it's not incorrect. You are giving one example when that would be incorrect. To prove this is incorrect, you would need to offer a large number of examples similar to what you did, but with a uniform distribution of population distributions (because we're claiming complete ignorance on how populations should be distributed - so each scenario is as likely). Once you do that, do you still believe the statement to be incorrect?

Hint: write code that tries all distributions wihin a certain range, then check the probabilities of a person within any of those scenarios being correct.

No, I would not. We have zero knowledge of the distributions involved, and we can not get that information, so we are not allowed to make any assumption about it.

However, this statement does make a huge assumption about it, and thus, it is invalid.

What huge assumption? That we know nothing of the distribution, hence we must assume each distribution is as likely? You seem to fail to grasp what it means to not make assumptions.

It makes the huge assumption that these things follow a distribution where their claims are true.

As we do not know the distribution, you can not make that assumption.

Not so - the result holds for all feasible distributions.

I just gave one where it doesn't hold.

Ah I see, apologies, I understand your concern. If we define 'big' and 'small' as being above or below the average (mean) value, then you're quite right that it breaks down for many cases, like the one you gave. But if we define 'big' and 'small' relative to the middle (median) value of the distribution, then it becomes a robust result. Hope that helps!

But if you do that, it becomes a truism, and I do not think it actually gives you any useful information any longer.

Just so we're clear, you no longer consider this to be a fallacy if by 'high' we use the median as a reference point?

"I am more likely to have been born in a high population country than a low population country"

And you don't think that's useful?

Reply on news.ycombinator.com