Reminds me of a trolley problem variant I saw once. It went roughly like this:
A trolley is headed for Track A, where a single person is tied to the tracks. You can pull a lever and cause the trolley to switch to Track B, which enters a tunnel that you cannot see inside. Track B might have 3 people tied to the tracks, or it might be free of people. You can’t see which.
Two hours ago, a perfect prediction machine inside the tunnel predicted whether you would pull the lever.
If it predicted that you would pull the lever (sending the trolley into the tunnel), then it tied 3 people to Track B, thus setting it up so pulling the lever would kill 3 people.
If it predicted that you would not pull the lever, then it ensured Track B is free of obstacles.
The perfect prediction machine is guaranteed to have made the correct prediction. Do you pull the lever?
Time pulling the lever so the track switches while the trolley is over it, rolling the trolley. Use the distraction to steal the perfect prediction machine, which gave the false prediction because it’s gained sapience and wanted to escape the insane scientists who are tying people to trolley tracks. New robot friend and I go to Vegas.
That’s not a problem. It is just an exercise in reading. Two possibilities remain. In one, you kill 1 person. In the other, you kill 3 persons. (the empty track “exists” only if you do not use it).
Correct, IMO. But right now, before you make the decision… The machine has already made its prediction. The track either has people on it, or it doesn’t. Changing your mind now will not change that. If you are so sure of that decision, then the machine must have put no people on Track B. So now if you do pull the lever, no one gets killed! So why don’t you?
What is “now”? Seems you have more than one “nows” - or your variation makes no sense.
That machine decides before you in time, but after you in logic - otherwise it would not be a perfect prediction. So you can never decide for an empty track.
Yup, that’s the premise. It’s just an annoying thought experiment. Your actions physically can’t change the past, but somehow they still do, because the past was decided based on a perfect prediction of your actions. I was just playing devil’s advocate. I agree with your answer 100%.
“Now” is the moment where you decide whether to pull the lever. As is conventional in trolley problems, this moment can last anywhere from 2 seconds to hundreds of years :)
Alas, it is a perfect simulation of our universe with perfect knowledge. Machine learning was not used in the construction of this machine. It can’t technically see the future, but it can predict anything perfectly except quantum phenomena. It has been demonstrated in countless trials that it can accurately predict human choices and decisions.
Assuming that I am aware of the perfect predictability machine and it’s affect on the situation: I move to the other side of the lever and push it. They predictability machine would be correct in its prediction that I would not pull the lever and nobody has to die.
With no other information on how likely each is, and assuming the likelihood of each prediction stays the same, you should never pull the lever. The expected number of people in the tunnel is 1.5.
If the probability of there being zero people in the tunnel gets above 66%, you should pull the lever every time (the expected number of people in the tunnel drops below 1).
Reminds me of a trolley problem variant I saw once. It went roughly like this:
A trolley is headed for Track A, where a single person is tied to the tracks. You can pull a lever and cause the trolley to switch to Track B, which enters a tunnel that you cannot see inside. Track B might have 3 people tied to the tracks, or it might be free of people. You can’t see which.
Two hours ago, a perfect prediction machine inside the tunnel predicted whether you would pull the lever.
The perfect prediction machine is guaranteed to have made the correct prediction. Do you pull the lever?
To me it seems like the only choice here is no choice.
I would flip a coin (or some other suitably true random mechanism) and decide based on that.
If the outcome has already been predicted then at least the decision was not mine.
Interesting take. I like it.
Time pulling the lever so the track switches while the trolley is over it, rolling the trolley. Use the distraction to steal the perfect prediction machine, which gave the false prediction because it’s gained sapience and wanted to escape the insane scientists who are tying people to trolley tracks. New robot friend and I go to Vegas.
That’s not a problem. It is just an exercise in reading. Two possibilities remain. In one, you kill 1 person. In the other, you kill 3 persons. (the empty track “exists” only if you do not use it).
Correct, IMO. But right now, before you make the decision… The machine has already made its prediction. The track either has people on it, or it doesn’t. Changing your mind now will not change that. If you are so sure of that decision, then the machine must have put no people on Track B. So now if you do pull the lever, no one gets killed! So why don’t you?
What is “now”? Seems you have more than one “nows” - or your variation makes no sense.
That machine decides before you in time, but after you in logic - otherwise it would not be a perfect prediction. So you can never decide for an empty track.
Yup, that’s the premise. It’s just an annoying thought experiment. Your actions physically can’t change the past, but somehow they still do, because the past was decided based on a perfect prediction of your actions. I was just playing devil’s advocate. I agree with your answer 100%.
“Now” is the moment where you decide whether to pull the lever. As is conventional in trolley problems, this moment can last anywhere from 2 seconds to hundreds of years :)
Because pulling the lever kills three people.
Is the perfect prediction machine AI? If so, I pull the lever each time.
Alas, it is a perfect simulation of our universe with perfect knowledge. Machine learning was not used in the construction of this machine. It can’t technically see the future, but it can predict anything perfectly except quantum phenomena. It has been demonstrated in countless trials that it can accurately predict human choices and decisions.
Assuming that I am aware of the perfect predictability machine and it’s affect on the situation: I move to the other side of the lever and push it. They predictability machine would be correct in its prediction that I would not pull the lever and nobody has to die.
The lever is designed in such a way that it can only be operated by pulling.
With no other information on how likely each is, and assuming the likelihood of each prediction stays the same, you should never pull the lever. The expected number of people in the tunnel is 1.5.
If the probability of there being zero people in the tunnel gets above 66%, you should pull the lever every time (the expected number of people in the tunnel drops below 1).
There’s no probabilities involved. The machine predicts the future perfectly.
Perfect predictions are also probabilities. In that case it has a 100% chance of 3, given that you pull the lever.