BOX A ONLY: 9

BOTH BOXES: 4

----------------------------------

Setup: You walk into a room. There are two boxes.

Box A (opaque) — contains either $1,000,000 or $0.

Box B (transparent) — contains $1,000 (you can see this).

A highly reliable, God-Like AI Predictor (almost always right) has already predicted, before you walked into the room, whether you will take only Box A or both boxes. Its prediction includes the assumption that you will be aware of the rules of the game and of it making this prediction.

If the Predictor predicted you will take only Box A, they put $1,000,000 in Box A.

If the Predictor predicted you will take both boxes, they put $0 in Box A.

Now the boxes are in front of you. You must choose right now, and your choice does not affect the Predictor anymore (the prediction and the filling already happened).

--------------------------------

Do you take only Box A (hoping for $1,000,000), or do you take both Box A and Box B (guaranteeing the $1,000 but possibly losing the million, but with a higher maximum payoff of $1,001,000)? Why?

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49731649)

what is your choice

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49731826)

id take both. the predictor would predict any rational person would take both, so it wouldn't matter, but its costless to see if it made an error since it's almost always right.

unless the predictor is in the first leg too (the predictor is predicited which move id take with a predictor). now im in a game of golden balls with essentially an ai clone. in which case id go with A, because $1000 is inconsequential and i think theres a greater than 1:1000 chance the predictor would think id choose that, so it's positive EV.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49731834)

if the predictor is predicting what'd id do in isolation, then ill take both

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49731865)

what exact same situation? isolation from what? huh?

the predictor is predicting (with 99.99999%+ accuracy) which choice you will make based on the rules of the game laid out in op. not sure if this helps or not

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49731909)

then i misred the question as "do you take A or B" instead of "do you take A or both?" under my A or B reading, the 100% correct answer would be B, and i thought you were asking this to see how many people would choose incorrectly because $1k is insignificant

if i had read your question correctly the context clue would have resolved the ambiguity. it was a rc fail on my part.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49731956)

although the way you are thinking about this is on the right track imo

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49731975)

i would choose A because $1k is inconsequential. im truly engaging in this hypo by *not* thinking of it further than that. if i choose A based on my gut, legitimately, not trying to game the predictor as it has already put/not put the money in A, then there's at least a chance the predictor would have predicted that (and under your hypo it *did* do that, because it was legitimately my first instinct.)

its kinda like quantum mechanics. the longer i think about it, the more likely i am to reach an optimal answer, which could be both. thus my actions right now do impact what the predictor has already done.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732051)

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49731790)

The rational choice for someone (a) who can think through the excercise and (b) for whom $1,000 is nbd is to take only Box A and hope the AI sees that you would see it that way.

The answer might change if the AI is predicting for a random, average person, who is more likely to take both boxes.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732114)

i thought about increasing the amount from $1000 but this is a famous philosophical problem so i didn't want to tweak it

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732144)

nm realized you probably want to know the name and background of it, i'll poast it later

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732188)

woah i was right about quantum retrocausality

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732203)

most poasters are very insecure about ever being wrong online. only like 15 people participated in my election guessing game where i offered a $50 giftcard. yet hundreds will make vague, non-falsiable "predictions."

from the face of the hypo most can tell its somewhat complicated and that there might be an objectively correct answer. and they'd rather not risk damage to their e-rep.

it's a lawyer forum, no suprise many are risk adverse bothboxmos

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732273)

(this is completely cr btw lmao all lawyers are the fucking same in this way)

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732287)

better call saul is the best legal show ever because it does a great job portraying this reality.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732312)

you can look up the problem to see what i mean

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732231)

the og hypo is from 1969, goy superstar should have made it $10m and $10k so that it's the same hypo adjusted for inflation.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732242)

this problem is more about exploring and explaining one's rationale (including assumptions) for one's choice than it is about the "correctness" of the choice itself

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732299)

let's say i offered to bet your entire net worth on something that was 51% in your favor paying even money. even though you have positive ev, you shouldn't do this, because the real cost of you losing everything isnt worth you having 2x. but if i offered you the same deal for $10, you would do it infinite times.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732331)

in your hypo, it is 100% clear, without any doubt whatsoever, that we would be risking our entire net worth if we made the bet

in the hypo in the OP, the *entire question* is whether or not one is actually "risking" anything if one chooses to take two boxes rather than just Box A. that is what it's all about

that is why it's important to fully engage with the hypo in the OP. because otherwise you're not giving a meaningful answer to the question

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732366)

"it's entirely rational to act how you want the AI to believe you would act in this scenario (therefore i choose only Box A)"

here you are filling in those assumptions. good. this is a totally fine answer. naturally though, a two-boxer is going to point out: "but the Predictor's decision has already been made. you lose nothing by taking the other box too; you just get a free +$1000 EV"

and then the thought experiment continues from here, because the real question is: under what assumptions/conditions *would* you actually be Risking the $1,000,0000 by taking both boxes?

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732406)

lets say im certain A is empty and twobox is the best move, but box B has a penny. i know im human and fallible, so even though im certain, id hedge on solipism because its almost costless. or reverse it and have box B have $999k. now, even if im nearly certain box A is full, the guaranteed money is too much to pass up.

the amount in the boxes, the deciders situation, and their confidence in their decision is all going to factor in.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732419)

it's a question pertaining to causality. if someone is caught up in thinking about the amounts of money involved, they are not thinking about the question hard enough, or are possibly just not intelligent enough to fully grasp the hypo. but i am pretty sure that everyone on this board is plenty smart enough to understand the depth of the hypo. it just takes some thinking, that's all

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732454)

still, a Smart Person should pretty easily recognize what the hypo is all about. it's just that answering it becomes very complex the more you get into it. seeing the dollar numbers and then immediately stopping there is - at the risk of Judging others - not what Smart People do imho

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732479)

if i wanted you to engage with something id frame it around something you actually care about. like when i tried to illustrate openmindedness to consuela. although that has its own pitfalls because sometimes people are so emotionally tethered to something they cant tolerate it being questioned. or they miss the point completely, like how consuela keeps saying im a bootlicking vaxxcuck.

ime smart people are even more susceptible to this than midwits. smart people tend to have a lot of ego about their intelligence and correctedness. e.g. risk adverse law fags terrified to risk being wrong even anonymously online about dumb shit.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732529)

i had a lol school prof who wrote articles about how judges often make illogical decisions, and to prove this he sent out a hypo like this with a correct answer that anyone could figure out, but most peoples gut reaction would be wrong. of course, he passed it out at some bar event where most people took two seconds and checked a box and moved on. this was published as if they had sat down and thought about it with the same intensity they would a matter before them!

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732630)

It's also totally unclear why someone would take both boxes or not, and if this has any effect on the chance the computer decides they were worthy or not of leaving the million.

The hypo is totally broken because it comes down to whether or not we trust the decision making ability of a computer, which may or may not be right or wrong, and can't even be proven to be so until after you've taken a box or boxes. The whole hypo suffers from a lack of clarity. Is it just a sword in a stone situation where we're just hoping the robot deems us worthy? I fail to see why there's any point in guessing what a robot thinks of us. Wouldn't it then follow that the optimal strategy would be to appeal to the robot's inner sense of goodness enough to leave the million in the first box? That's what we're left with ultimately, and, perhaps, unfortunately.

If you're insinuating that the best course is to take Box B, prove the computer right about the goodness of humanity so that it must leave the million, then build a time machine and go back in time to take the first box, then I think you could have articulated that a lot better.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732292)

Right so it's a shitty hypo and it depends on interpretation to arrive at the optimal strategy.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732436)

https://en.wikipedia.org/wiki/Newcomb%27s_problem

the real meat of the problem is getting into causality and whether or not "free will" is a real thing, and under what conditions it is/could be real

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49732611)

The point though, is that there is no way for the predictor to "almost always" reach the correct decision unless it is able to analyze the specific decisionmaker and replicate their thought process with a high degree of accuracy. Being correct almost always would involve correctly predicting how a wide range of intellects would approach the problem.

So, there is no probabilistic "correct" answer between one box and two boxes under this hypothetical. The actual analysis of the choice itself is a red herring. The nature of the predictive device and how it works is the most important part of the hypo, and my interpretation is that you maximize your outcome by making whatever choice is consistent with your internal reasoning process. Or alternatively, if the predictor is even able to predict that you might disregard your internal reasoning and make the opposite choice, then it's simply impossible to lose the hypo.

Either way, the choice is irrelevant. However, the hypo is somewhat incomplete because it does not describe the circumstances under which the predictor will be inaccurate, which could entirely change the problem.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733205)

OP adds in the term "godlike" to the classic problem, which was the source of my assumption about how the box might work, but it doesn't appear to be part of the classic problem, which is more wide open. What is interesting though is that it seems like even many of the lines of reasoning that correctly focus on the nature of the predictor are still getting caught up trying to prove whether one box or two boxes is "correct."

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733294)

that's the whole point of the problem

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733305)

-----------------

In his 1969 article, Nozick noted that "To almost everyone, it is perfectly clear and obvious what should be done. The difficulty is that these people seem to divide almost evenly on the problem, with large numbers thinking that the opposing half is just being silly."[1][4] The problem continues to divide philosophers today.[9][10] In a 2020 survey, a modest plurality of professional philosophers chose to take both boxes (39.0% versus 31.2%).

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733320)

the reason why people are split down the middle about their answer to this problem is that people make different instinctive assumptions about the predictor. it's interesting to explore why this is. that is the significance of the problem

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733333)

The reason this problem is so popular is it's intentionally vague. You can claim it's interesting to see how much people assume about a predictor whose accuracy is unknowable, or you could take a much more scientific approach and actually decide on the parameters vs make people guess what they are. If you think you can draw meaningful inferences based on how people make their inferences, based on assumptions about an unknowable variable, then I would say you haven't learned very much at all.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733353)

that is definitely an important assumption that must be made as part of the problem

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733706)

--------------

In the standard version of Newcomb's problem, two boxes are designated A and B. The player is given a choice between taking only box B or taking both boxes A and B. The player knows the following:[4]

Box A is transparent, or open, and always contains a visible $1,000.

Box B is opaque, or closed, and its content has already been set by the predictor:

If the predictor has predicted that the player will take both boxes A and B, then box B contains nothing.

If the predictor has predicted that the player will take only box B, then box B contains $1,000,000.

The player does not know what the predictor predicted or what box B contains while making the choice.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733778)

simple as. i'm a simple goy who doesn't believe in nonsensical "jewish physics" like retrocausality

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733679)

Also in order to one box you have to believe in either retrocausality or a non-causal decision theory, both of which are incredibly Jewish fwiw

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733761)

simple-hearted yet not even in the slightest simple-minded (even your [cyber]detractors can't deny your formidable mental horsepower, IIRC, but then again Talmudic rhetoric is 180 so who knows), or:

Wise as a serpent yet innocent as a dove tp :-)

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733732)

retrocasuality isn't necessary unless you abstract it to multiple timelines or something. you can only make the decision once, and the predictor knew what you'd make, so the money was already there or not. your decision didn't cause things to change in the past.

lets say every morning you decided between coffee and tea, but always choose coffee when it's raining, but you weren't aware you had this tell. the next rainy day i boasted i could predict your choice, wrote coffee on a piece of paper face down, and let you choose. you still have freedom of choice, but you just don't know i know what you're going to do. even though you *could* have chosen tea, i know you won't because im aware of a habit you don't even recognize about yourself. and you can only make the decision *once*. the trap is thinking "but I could have chosen tea!" yes, you *could* have, you do have free will, but there's one timeline, one discrete binary event that only occurs once.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734496)

If the computer has the potential to be wrong then it is a capricious god and you should take two boxes every time.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733314)

This intriguing thematic premise must have been a classic SciFi movie or TV episode or anime and/or Japanese video game or Autoadmit performance art or something at some point, right? Urgently in need of thought provoking cultural curation, Thank.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49733740)

and before anyone says "yeah but its impossible!", why dont don't they have engaging with other impossible hypos like "what if you were immortal" or "what if you could turn into a ghost." especially in 2026 when algorithems can already very accurate predict things about a person.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734405)

This is why the predictor has to have a chance to be wrong for the hypo to be a hypo worth considering

I always take two boxes because retrocausality is not possible. One's decisions cannot change the past. If you take both boxes and there is no million dollars in box A, all that means is that the million dollars was never an option for you to begin with. It's not due to any decision you made

There is a branch of decision theory that makes a pretty good argument for one boxing. Functional decision theory. It's very Jewish but it does make sense in the context of this particular problem. But if you try to apply it to all decisions it starts not making sense. I personally just don't buy it

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734557)

You’re not “risking” anything or “betting that the predictor is wrong,” because your decision cannot change the past decision that the predictor already made. You’re just taking a free $1000 more than you’d otherwise get in all situations

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734578)

you've decided that if it works that implies something impossible, and so you're going with a decision you think is best for in reality as you understand it, which is different from the hypothetical one.

and the 99.9999% vs 100% is insignificant anyway from a mathmatical decision making POV. take the predictor out of the equation -- what if i offered you the choice between $1m if an RNG doesn't generate 1 out of trillion numbers or $1m if the RNG hits + $1k? the EV of option 1 is $1m minus a tiny fraction of a cent, the EV of option 2 is $1k plus that fraction.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734627)

The hypo is moot if the predictor is necessarily always correct no matter what. The hypo then becomes: "If you take only Box A, it will necessarily contain one million dollars." It's moot

The hypo you wrote out doesn't have anything to do with this problem because it is missing the meaningful part of the problem

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734652)

you're thinking about my question in terms of the actual purpose of the hypo. im asking you why your decision would change if it was 99.9999% versus 100%, which has nothing to do with it. thats why im framing it with probablity for an event occurring after your decision than one that happens before.

if you accept that the predictor can predict with 99.999% accuracy, then the odds of it being wrong are the same as the RNG occurring after. mathematically choosing the 99.999% outcome to win a million is best.

if not, then explain why your decision for one 99.99% event is different from the other? as far as i can tell, the only difference is that you don't believe the predictor is possible, i.e. fighting your own hypo.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734713)

You are smart and understand the hypo so I know that you're capable of clearly communicating your ideas about it. You're doing the same thing that hatp does and being lazy and just writing shit out stream of consciousness style instead of spending a few more seconds editing your posts as needed to make sure you're communicating clearly

I already answered what (I think) you're asking. You actually should put your questions to grok or another LLM. They are very good at explaining these sorts of problems that have a lot of written human analysis about them. It will also explain under which assumptions and decision theories it makes sense to one box

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734737)

i will try to explain again and if its still unclear ill copy and paste into grok when i get home. im also gonna use 99% but i mean 1 in a trillion and don't want to clutter with repeating 9s:

if you accept the AI is 99% accurate it is the same as the RNG for the purpose of the hypo. the timing doesnt matter either, the RNG event could have happened before your decision with the results unknown to you, and the decision calulus is the same.

so, without getting into the mechanics of the event/prediction (what you think it implies about reality, whether you think its possible, casualty, etc), then why does 99% versus 100% impact your decision? because any dispute over the mechanics of *how* predicted at 99% is fighting your own hypo. you must accept that it is 99%, which is exactly the same my RNG analogy.

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734811)

If you treat Newcomb's Paradox as strictly an EV calculation, like you are, then one boxing is correct. And some decision theories do this. Just ask an LLM. Also ask it about the Parfit’s Hitchhiker hypo. I think it is actually a more interesting version of this problem and it's definitely stronger evidence against causal decision theory

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734836)

im asking why your decision is different if the event has a 99% chance of occurring versus 100%. i understand the purpose of the hypo and how being certain or near certain impacts that. this doesn't resolve why your decision changes other than the reason you've stated, which is fighting your own hypo (retrocausality doesnt exist therefore the decision is made therefore its empty/full therefore take the $1k. see? i get it.)

your decision shouldnt change if you're not fighting the hypo. you could just say "an event has a 99% chance of occurring." if you accept that, then your decision shouldn't fhange. if you think that this *specific* event *must* require retrocasuality, then under the hypo it does. so if ylu say "but retrocasuality doesn't exist" that is the same as "but i did have breakfast this morning."

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734867)

GS keeps saying but you CAN'T effect the decision. This is true. But also computer's can't predict human behavior perfectly either. So there's a lot of 'cant's' in this one.

The whole exercise is a Rorschach test with no correct answer and you can see it however you want to see it. But an infallible or near infallible computer plus humans making choices with free will are incompatible. It's no different than saying ok God has chosen you for salvation, do you sin or never sin? Goy Superstar is saying that he's gonna sin either way because his salvation is predestined. This is actually what the Calvinists decided about this problem too. They decided if you were in the club, you could fuck up and do bad things but you were still in the club by virtue of being chosen. They called it Preservation of the Saints. If you're already good in God's book you could do no wrong. That's how he's playing the game. Sin because it's all worked out for you (or not). Pretty cynical behavior imo

(http://www.autoadmit.com/thread.php?thread_id=5843901&forum_id=2ГѓЖ’#49734739)