Hehe thats why I think the original question of Box A being $1000 and Box B being a million was kinda boring, since $1000 is barely anything in today's world. 3 more zeroes does making things more interesting
I’d much rather take a sure million with a (slight?) chance of a bonus billion, versus an unknown chance at 0 or a billion. I could do plenty with a million that would significantly change my life for the better.
But I would probably do the opposite if A contained $1000 and B contained a potential million as in the original example. $1000 is a tolerable amount to risk missing out on.
If I wanted to use logic, I'd say taking both A and B is the only way to have a guaranteed $1,000,000 outcome, because B only could get you money but also nothing.
But, if I choose B only, I'm sort of "forcing" the machine into that kind of prediction, right? I don't know about this experiment, but since your post says it's a paradox, I think that's how it works.
So my choice is B, the machine has predicted it and I get a nice $1,000,000,000.
ka = the potential prize in box B; i.e. "k times larger than a"
p = the odds of a false positive. That is, the odds that you pick box B only and it got nothing, because dumb machine assumed that you’d pick A too.
n = the odds of a false negative. That is, the odds that you pick A+B and you get the prize in B, because the machine thought that you wouldn’t pick A.
So the output table for all your choices would be:
pick nothing: 0
pick A: a
pick B: (1-p)ka
pick A+B: a + nka
Alternative 4 supersedes 1 and 2, so the only real choice is between 3 (pick B) or 4 (pick A+B).
You should pick A+B if a + nka > (1-p)ka. This is a bit messy, so let’s say that the odds of a false positive are the same as the odds of a false negative; that is, n=p. So we can simplify the inequation into
a + nka > (1-n)ka // subbing “p” with "n"
1 + nk > (1-n)k // divided everything by a
1 + nk - (1-n)k > 0 // changed sides of a term
1 + 2nk -k > 0 // some cleaning
n > (k-1)/2k // isolating the junk constant
In OP’s example, k=1000, so n > (1000-1)/(2*1000) → n > 999/2000 → n > 49.95%.
So you should always pick B. And additionally, pick A if the odds that the machine is wrong are higher than 49.95%; otherwise just B.
Note that 49.95% is really close to 50% (a coin toss), so we’re actually dealing with a machine that can actually predict the future somewhat reliably, n should be way lower, so you’re probably better off picking B and ignoring A.
I meant, let’s imagine the machine predicted B and is wrong (because I take A+B). I would call that scenario „I have free will - no determinism.“ Then I will have 1.000.000.000 „only“. That’s a good result.
Except that's not the worst case. If the machine predicted you would pick A&B, then B contains nothing, so if you then only picked B (i.e. the machine's prediction was wrong), then you get zero. THAT'S the worst case. The question doesn't assume the machine's predictions are correct.
I think the major unanswered question is how reliable do we think the machine is? 50%? 100%? I think the most interesting scenario is one where we are convinced that the machine actually predicts the future and always predicts correctly, so I’ll continue with that assumption in mind.
From one point of view, we have no reason not to take both boxes, since we can’t alter the machine’s prediction now, it’s already happened. I think however that this undermines my premise. Choosing both boxes only makes sense if we don’t actually believe the machine predicts the future.
One would be tempted to say "alright, then I will choose only box B, as the machine will have predicted that and I will get lots of money. If I were to choose both boxes, the machine would have predicted that too, and I would get much less money.
My argument is that both answers are wrong in a sneaky way: assuming an actual perfect predictor, my answer is box B only. However, the important part here is that this will not be, in fact, a choice. The result was already determined ahead of time, so I really only had that one option.
You never know the shenanigans of a machine, and one million is largely enough for me until I die, or if science gives us the option to live forever I bet machines will do the work for us :-)
Edit: as I believe the machine can be wrong, I’d probably take A + B
Here’s my solution to Newcomb’s Paradox: the predictor can be perfectly infallible if it records your physical state and then runs a simulation to predict which box you’ll pick. E.g. it could run a fancy MRI on you as you are walking through the hallway towards the room, quickly run a faster-than-real-time physical simulation, and deposit the correct opaque box into the room before you open the door. The box, the hallway, the room, the door are all part of the simulation.
Here’s the thing: a computer simulation of a person is just as conscious as a physical person, for all intents of “consciousness”. So as you are inside the room making your decision, you have no way of knowing if you are the physical you or the simulated you. The predictor is a liar in a way. The predictor is telling the simulated you that you’ll get a billion dollars, but stating the rules is just part of the simulation! The simulated you will actually be killed/shut down when you open the box. Only the physical you has a real chance to get a billion dollars. The predictor is counting on you to not call it out on its lie or split hairs and just take the money.
So if you think you might be in a simulation, the question is: are you generous enough towards your identical physical copy from 1 second ago to cooperate and one-box? Or are you going to spitefully deprive them of a billion dollars by two-boxing just because you are about to be killed anyway? Remember, you don’t even know which one you are. And if you are the spiteful kind, consider that we are already making much smaller time-cooperative trade-offs all the time, such as the you-now taking a breath just so that the you-five-seconds-from-now doesn’t suffocate to death.
What if the predictor doesn’t use a MRI or whatever? I posit that whatever prediction method it uses, if the method is sufficiently advanced to be infallible then somewhere in the process it MUST be creating conscious observer instances.
Box A and B as the prediction has already been made so the choice has no bearing on the contents at this point. You either get the guaranteed million or both.
Well what you choose may not direct affect what is inside Box B, but there is still a huge difference between the two choices.
Imagine the way that the machine did it's prediction was copying your brain and making this copied brain choose in a simulation. Assuming the copied brain is completely identical to your brain, the machine could predict with 100% accuracy what the real you would choose. In this sense, what you choose can affect what's inside Box B (or rather, what your copied brain chooses can affect whats inside Box B).
One more thing to think about: How do you know that you aren't the simulated brain that's been copied?
I feel like unless we're talking about supernatural AI the only answer is A&B
Otherwise the box has no real way of knowing what you would've picked, so it's complete RNG.
If there was a realistic way that it could make that decision i'd choose only B, but otherwise it just doesn't make sense.
edit: I also didn't realize until after I read it that box A always has the million dollars. So there's actually no reason to pick only box B in this scenario. The paradox only makes sense if box A is significantly less than box B. It's supposed to be a gambling problem but A&B is completely safe with the changes made.
The machine has already done it's prediction and the contents of box B has already been set. Which box/boxes do you take?
If my choices don't matter and the boxes are predetermined, what point is there to only taking one box? The machine already made its choice and filled the boxes, so taking both boxes is always the correct answer. Either I get $1,000,000 if the machine thought I would take both, or I get $1,001,000,000 if it didn't. This is a false dilemma, there is never a reason to take just one box.
This isn't a false dillemma. Imagine if the way the machine predicts is by copying your brain and putting it in a simulated reality, then the copy of you gets asked to choose which boxes to take, the exact same way and be given the exact same information. Under this assumption, the machine could predict with 100% accuracy what the real you would've chosen.
How do you know you are even the real you. You could just be the machine's simulation of the real you.
There is a dilemma and the dilemma is about how much you want to trust the machine.
If you are a simulation, then your choice doesn't matter. You will never get any real benefit from the boxes. It's like saying, "there is also a finite possibility that the machine is lying and all the boxes are empty". In which case, the choice is again irrelevant.
Situations in which your choice doesn't matter are not worth considering. Only the remaining possibility, that you are not a simulation and the machine is not lying, is worth considering.
Add comment