I was really glad to see you take into account albedo changes-I think this is really important for the boreal forest.
For the existential and overall cost effectivenesses, I think youâre getting implausible values because the standard deviation of your future population size and duration of humanity are too large. For instance, it appears that your 5th percentile of future population of humanity is 10^-34 people, and duration is <1 yr. I guess itâs possible to have a mean greater than the 95th percentile, but I donât think the means are plausible. Anders Sandberg estimates the upper bound of number of human lives is around a Googol (10^100, note, Google is a misspelling of Googol). So it is not reasonable that 1 ÂŁ is going to save 10^90 lives.
For the existential and overall cost effectivenesses, I think youâre getting implausible values because the standard deviation of your future population size and duration of humanity are too large.
You may well be right. I did not try to get the most accurate distributions. The selected ones are essentially for illustration that the longterm effect of tree planting dominates the analysis.
In the table of this section, I mention that such effect is of the order of 10^(10^2) in terms of lives saved per hectare. This was also to indicate I am only confident of getting the order of magnitude of the order of magnitude right. So, for the cost-effectiveness, I believe log10(log10(âexistential/âoverall cost-effectiveness (life/âÂŁ)â)) = 2.
For instance, it appears that your 5th percentile of future population of humanity is 10^-34 people, and duration is <1 yr.
The virtually null 5th percentile of the future population size makes sense to me. Given Toby Ordâs guess if 1â2 for the total existential risk, I guess the median population size is close to 0 (I set it to 1). The 5th percentile of the humanityâs lifespan being shorter than 1 year does not make sense, but it does not affect the mean of the distribution, which is driven by the right tail. In general, truncated distributions would be more accurate.
Okay-if youâre just trying to get the order of magnitude of the exponent correct, then this could be consistent with 10^10^1.5 ~10^30 lives/âÂŁ, which is plausible. As for the median population size, I guess you are trying to get an average over the time horizon which could extend 100 trillion years, so thatâs why youâre saying itâs around zero with 50% as existential risk, which makes sense.
By the way, I was confused when I read this:
The parameters of humanityâs future duration were estimated as follows:
I think you mean population size instead of duration.
As for the median population size, I guess you are trying to get an average over the time horizon which could extend 100 trillion years, so thatâs why youâre saying itâs around zero with 50% as existential risk, which makes sense.
Yes, that was it, thanks for clarifying!
I think you mean population size instead of duration.
I was really glad to see you take into account albedo changes-I think this is really important for the boreal forest.
For the existential and overall cost effectivenesses, I think youâre getting implausible values because the standard deviation of your future population size and duration of humanity are too large. For instance, it appears that your 5th percentile of future population of humanity is 10^-34 people, and duration is <1 yr. I guess itâs possible to have a mean greater than the 95th percentile, but I donât think the means are plausible. Anders Sandberg estimates the upper bound of number of human lives is around a Googol (10^100, note, Google is a misspelling of Googol). So it is not reasonable that 1 ÂŁ is going to save 10^90 lives.
Hi David,
Thanks for the comment!
You may well be right. I did not try to get the most accurate distributions. The selected ones are essentially for illustration that the longterm effect of tree planting dominates the analysis.
In the table of this section, I mention that such effect is of the order of 10^(10^2) in terms of lives saved per hectare. This was also to indicate I am only confident of getting the order of magnitude of the order of magnitude right. So, for the cost-effectiveness, I believe log10(log10(âexistential/âoverall cost-effectiveness (life/âÂŁ)â)) = 2.
The virtually null 5th percentile of the future population size makes sense to me. Given Toby Ordâs guess if 1â2 for the total existential risk, I guess the median population size is close to 0 (I set it to 1). The 5th percentile of the humanityâs lifespan being shorter than 1 year does not make sense, but it does not affect the mean of the distribution, which is driven by the right tail. In general, truncated distributions would be more accurate.
Okay-if youâre just trying to get the order of magnitude of the exponent correct, then this could be consistent with 10^10^1.5 ~10^30 lives/âÂŁ, which is plausible. As for the median population size, I guess you are trying to get an average over the time horizon which could extend 100 trillion years, so thatâs why youâre saying itâs around zero with 50% as existential risk, which makes sense.
By the way, I was confused when I read this:
I think you mean population size instead of duration.
Yes, that was it, thanks for clarifying!
Good catch, corrected!