When you say “the year of PASTA”, you probably mean the year than AI appears with 50 per cent probability. But why “50 per cent probability”? 10 per cent seems to be more important. For example, when we say “human life expectancy is 75 years”, it means that in the half of the worlds I will die before 75. The same way, by using the median year as a measure of AI timing, you already accept the loss of the half of human future when AI will appear before that date.
More generally, speaking about the “year of AI” is meaningful only if the dispersion of the Probability-of-AI-appearance(t) is small. If 10 per cent is 2030, 50 per cent is in 2100 and 90 per cent is in the year 3000, than saying that AI will appear in 2100 is a completely misleading picture.
That is, there are two problem in using year as a way to estimate AI-timing: 1) humanity will go extinct in the half of cases before this year 2) it creates a false impression that AI probability of appearance is a bell-like curve with a small deviation from the mean.
when we say “human life expectancy is 75 years”, it means that in the half of the worlds I will die before 75
Life expectancy is the mean value of the distribution, not the median value. Half of the worlds are below the median value, which is not life expectancy.
Yes. I found that “male life expectancy at birth is 76 years. That’s the mean value. The median life expectancy is just past age 80. And the mode (i..e, most common) age at death is age 86.”
When you say “the year of PASTA”, you probably mean the year than AI appears with 50 per cent probability. But why “50 per cent probability”? 10 per cent seems to be more important. For example, when we say “human life expectancy is 75 years”, it means that in the half of the worlds I will die before 75. The same way, by using the median year as a measure of AI timing, you already accept the loss of the half of human future when AI will appear before that date.
More generally, speaking about the “year of AI” is meaningful only if the dispersion of the Probability-of-AI-appearance(t) is small. If 10 per cent is 2030, 50 per cent is in 2100 and 90 per cent is in the year 3000, than saying that AI will appear in 2100 is a completely misleading picture.
That is, there are two problem in using year as a way to estimate AI-timing: 1) humanity will go extinct in the half of cases before this year 2) it creates a false impression that AI probability of appearance is a bell-like curve with a small deviation from the mean.
Life expectancy is the mean value of the distribution, not the median value. Half of the worlds are below the median value, which is not life expectancy.
Yes. I found that “male life expectancy at birth is 76 years. That’s the mean value. The median life expectancy is just past age 80. And the mode (i..e, most common) age at death is age 86.”