The Nonlinear Library: LessWrong LW - Datasets that change the odds you exist by dynomight
Jun 30, 2024
10:07
Link to original article
Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Datasets that change the odds you exist, published by dynomight on June 30, 2024 on LessWrong.
1.
It's October 1962. The Cuban missile crisis just happened, thankfully without apocalyptic nuclear war. But still:
Apocalyptic nuclear war easily could have happened.
Crises as serious as the Cuban missile crisis clearly aren't that rare, since one just happened.
You estimate (like President Kennedy) that there was a 25% chance the Cuban missile crisis could have escalated to nuclear war. And you estimate that there's a 4% chance of an equally severe crisis happening each year (around 4 per century).
Put together, these numbers suggest there's a 1% chance that each year might bring nuclear war. Small but terrifying.
But then 62 years tick by without nuclear war. If a button has a 1% chance of activating and you press it 62 times, the odds are almost 50/50 that it would activate. So should you revise your estimate to something lower than 1%?
2.
There are two schools of thought. The first school reasons as follows:
Call the yearly chance of nuclear war W.
This W is a "hidden variable". You can't observe it but you can make a guess.
But the higher W is, the less likely that you'd survive 62 years without nuclear war.
So after 62 years, higher values of W are less plausible than they were before, and lower values more plausible. So you should lower your best estimate of W.
Meanwhile, the second school reasons like this:
Wait, wait, wait - hold on.
If there had been nuclear war, you wouldn't be here to calculate these probabilities.
It can't be right to use data when the data can only ever pull you in one direction.
So you should ignore the data. Or at least give it much less weight.
Who's right?
3.
Here's another scenario:
Say there's a universe. In this universe, there are lots of planets. On each planet there's some probability that life will evolve and become conscious and notice that it exists. You're not sure what that probability is, but your best guess is that it's really small.
But hey, wait a second, you're a life-form on a planet with conscious life! Given that you exist, should you increase your guess for how likely conscious life is to evolve on a random planet?
Again, you have two schools of thought. One says yes, you have data, increase your guess, while the other says no, don't increase, if there wasn't life you wouldn't be here, anthropic principle - anthropic principle!
4.
After many years of being confused by these questions, I think I now understand what's happening. These questions are confusing because they're actually about a sort of narrow technical question, and only appear to be about to the fact that you might not exist.
To explain, let me introduce another scenario:
One day you wake up at my house. As you groggily look around, I explain that you've been invited to Dynomight family dinner! And that the way that DFD works is:
1. I sneak into your house at night, anesthetize you, and bring you to my lair.
2. When you wake up, I make you some delicious Fagioli all'Uccelletto.
3. After you've eaten, I bring out a box containing a bunch of identical revolvers. Half have no bullets in them, while the other half have bullets in all six chambers. You pick one revolver at random, put it to your head, and pull the trigger. (To refuse would be a huge faux pas.)
4. If you're still alive, I bring out a $100 bill and offer to sell it to you for $60. If you agree, I take your gun and see if it has bullets in it. If it's empty, then I take your $60, give you the $100, and ask you to come back soon. If not, I take your $60 but don't give you the $100, welcome to dinner at my house, chump.
So you eat the Fagioli all'Uccelletto (it is excellent) and you play the mandatory revolver game and don't die, and I offer you the $100. Should you accept?
Yes you should. There's ...
Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Datasets that change the odds you exist, published by dynomight on June 30, 2024 on LessWrong.
1.
It's October 1962. The Cuban missile crisis just happened, thankfully without apocalyptic nuclear war. But still:
Apocalyptic nuclear war easily could have happened.
Crises as serious as the Cuban missile crisis clearly aren't that rare, since one just happened.
You estimate (like President Kennedy) that there was a 25% chance the Cuban missile crisis could have escalated to nuclear war. And you estimate that there's a 4% chance of an equally severe crisis happening each year (around 4 per century).
Put together, these numbers suggest there's a 1% chance that each year might bring nuclear war. Small but terrifying.
But then 62 years tick by without nuclear war. If a button has a 1% chance of activating and you press it 62 times, the odds are almost 50/50 that it would activate. So should you revise your estimate to something lower than 1%?
2.
There are two schools of thought. The first school reasons as follows:
Call the yearly chance of nuclear war W.
This W is a "hidden variable". You can't observe it but you can make a guess.
But the higher W is, the less likely that you'd survive 62 years without nuclear war.
So after 62 years, higher values of W are less plausible than they were before, and lower values more plausible. So you should lower your best estimate of W.
Meanwhile, the second school reasons like this:
Wait, wait, wait - hold on.
If there had been nuclear war, you wouldn't be here to calculate these probabilities.
It can't be right to use data when the data can only ever pull you in one direction.
So you should ignore the data. Or at least give it much less weight.
Who's right?
3.
Here's another scenario:
Say there's a universe. In this universe, there are lots of planets. On each planet there's some probability that life will evolve and become conscious and notice that it exists. You're not sure what that probability is, but your best guess is that it's really small.
But hey, wait a second, you're a life-form on a planet with conscious life! Given that you exist, should you increase your guess for how likely conscious life is to evolve on a random planet?
Again, you have two schools of thought. One says yes, you have data, increase your guess, while the other says no, don't increase, if there wasn't life you wouldn't be here, anthropic principle - anthropic principle!
4.
After many years of being confused by these questions, I think I now understand what's happening. These questions are confusing because they're actually about a sort of narrow technical question, and only appear to be about to the fact that you might not exist.
To explain, let me introduce another scenario:
One day you wake up at my house. As you groggily look around, I explain that you've been invited to Dynomight family dinner! And that the way that DFD works is:
1. I sneak into your house at night, anesthetize you, and bring you to my lair.
2. When you wake up, I make you some delicious Fagioli all'Uccelletto.
3. After you've eaten, I bring out a box containing a bunch of identical revolvers. Half have no bullets in them, while the other half have bullets in all six chambers. You pick one revolver at random, put it to your head, and pull the trigger. (To refuse would be a huge faux pas.)
4. If you're still alive, I bring out a $100 bill and offer to sell it to you for $60. If you agree, I take your gun and see if it has bullets in it. If it's empty, then I take your $60, give you the $100, and ask you to come back soon. If not, I take your $60 but don't give you the $100, welcome to dinner at my house, chump.
So you eat the Fagioli all'Uccelletto (it is excellent) and you play the mandatory revolver game and don't die, and I offer you the $100. Should you accept?
Yes you should. There's ...