gotta save women is 3,344 words long and tagged [Adventure] [Human] [Main6]

From LessWrong:

Signaling is defined by Yvain as "a method of conveying information among not-necessarily-trustworthy parties by performing an action which is more likely or less costly if the information is true than if it is not true". Some signaling is performed exclusively to impress others (to improve your status), and in some cases isn't even worth that. In other cases, signaling is a side-effect of an otherwise useful activity.

For example, if doing something is easy for one type of person and hard for another type of person, you might do that thing just to get people to think you're the former type of person, even if the thing isn't in itself worth doing. This could explain many facets of human behavior, [such as buying expensive jewelry (social signaling) or pursuing ineffective habits for appearance] and reveal opportunities for reducing waste.

Not all signaling is about abilities. Signaling can also be about personality, current emotional state, beliefs, loyalty to a particular group, status within a group, etc.

The degree to which Signaling influences human behavior is hard to overstate. Every day, countless conflicts break out because of Signaling. It is almost certainly the reason why Raw Cringe does what he does on FimFiction. So-called "trolling" is likely only rarely motivated by a greater cause, and otherwise also a means for signaling. If you didn't quite get what it is, go back and re-read the definition, or even do a bit of research – any time used to read about signaling is almost certainly time well spent.

Raw Cringe very clearly thinks of himself as operating on some level of irony. In my experience, people in his kind of position usually don't have coherent utility functions or even coherent goals or rules. It is probable that there is no fixed number of irony layers, no grand plan behind his actions, no clear FimFiction persona that is distinctly different from his real character. He is likely operating primarily on instinct, and doesn't even come up with clever justifications after the fact, let alone before.

If someone implicitly pretends to be acting on at least one layer of irony, is she more likely to support or not to support what she says? I'm fairly confident that the correct answer is the former. If I hated women, I wouldn't go and write stories with a balanced cast and a realistic portrayal of female characters, only to then think "ha, ha, it is was only ironic, I think women are all stupid, I was playing three-dimensional chess, GOTCHA!" No, even if I was less subtle, it is a mistake to assume that others think deeply enough about what I do to get it, and this is not a mistake I am likely to make. And at the end of the day, I want to degrade women, not empower them.

This is not a random example. I've done a bit of research into R/C's blog, and one consistently recurring subject is the female gender. (Not necessarily the only subject, though, others are Anime, Yu-Gi-Oh and, oddly, god.)

And the theme that is consistent when talking about gender is objectification. Not so much shaming or degrading (at least not always), but clear objectification, seemingly from a point of unfulfilled desire, honest ignorance, and dwelt-up frustration; a conclusion that sounds boring, and which is probably thrown around more often than it is should, but one that nonetheless seems to be true here,

Raw Cringe, this I am almost certain of, has above average intelligence. In fact, I'd bet a lot of money that he is between 110 and 130. You could say a lot of things about his writing, but it is not gibberish. There is cohesion. And there also is self-awareness, not to the point of being strategic and acting on it, but enough to recognize that his writing is objectionable, and to deliberately play it up at parts.

To illustrate this – and if you got this far you should be interested enough to take this journey a bit further – I present to you this blogpost:

All right, fimfiction, confession time: I really, really dislike 99% of modern women.
Don't get me wrong. I'm not saying that the female of the species is inherently bad. I'm also not saying that I hate all women. All I'm saying, is just that I personally can't stand almost any individual female humans. No woman has ever improved my life by her presence in it. Let me give you a few examples:
Example #1: My stepmom
My dad married a woman who left us while I was still an infant. I don't remember anything about her except a lot of screaming. I do remember that one time she took me grocery shopping when I was about two years old, and she had me in the baby seat in the cart. I guess I misbehaved because she pretended to abandon me in the middle of the grocery store aisle... she walked away and was out of my sight for at least five or ten minutes, and I couldn't get out of the seat.
I believe this woman's actions, and in particular this incident. is certainly a contributing if not a causal factor in the crushing sense of separation anxiety I feel to this day. Science shows that psychological trauma performed upon infants has echoing effects that persist with a child throughout his adult life. If you don't believe me, just google "MK Ultra brainwashing" and have your mind blown. I think a lot of regular people, people with average intelligence, aren't necessarily aware of the profound psychological impacts that early-childhood and formative-year parental actions (or rather mis-actions) can have on a developing human being, and that this ignorance and lack of scientific literacy is a serious political issue in terms of what kind of parental rights that females should legally be granted in case of divorce, separation, etc... FACT, in 127 of 392 United States jurisdictions a man can lose parental visiting rights just for catching the flu... FACT, Men are fourteen times as likely to be raped in prison than women... And when it comes to abortion, the father never gets a say, most judges will laugh you out of the courtroom if not lock you up for making the straightforward legal argument that a man ought to have the same right to control his own reproduction that a woman does, and that just because a woman wants to bear the child to term does not take away the father's implicit legal right to force an abortion... the point is that modern western FEMINIZED government is built on a series of concentric stone rings which are the respective cornerstones of none other than FEMINISM, a.k.a. the doctrine of female supremacy... and then your average sluttily-dressed bitch on the street, wearing low-cut blouses with no bra, you would be lucky if she is even wearing panties, and this lack of civil virtue and decency began its descent in earnest with the "flappers" of the 20's and perhaps even earlier, I have read books that trace the root of feminism back through Marx and Hegel all the way to the heresiarch himself, Martin Luther... and women are so stupid, they always go for the big hulking assholes who beat them up, and they make excuses for them and say things like, "I know he really loves me deep down, he just doesn't know how to express it," and, "He'll change one day, I believe in him... He promised me he'll do his best to change," you know, lines that show that their internal evolutionarily programmed rationalization engine is firing on all cylinders—here's another FACT, by the way: did you know that 76% of women who report being raped also report experiencing orgasm during the event? This is clear-cut, irrefutable evidence that the primary motive basis of female sexuality is rooted by evolutionary forces in hypergamy (i.e. leaving your current man as soon as you find a man more powerful), which accounts for their notorious irrationality, refusal to look at evidence or listen to reasoned arguments, as well as their inherent adulterous nature (women in America initiate divorce at a much higher rate than men)... if the next tribe over makes war on yours and kills all the men, you as a woman stay alive by fucking the guy who killed your old husband... I have personal anecdotal experience of this phenomenon myself, when this one total bitch, my buddy's girlfriend in high school, she liked to lead me on just to make him jealous...
Anyway, I'm basically done with 3DPD women (that's "three-dimensional pig, disgusting", in case you didn't know). I'm going to stick with anime girls from now on.
Anime is so great. (Women never appreciate anime, they only like "soap operas", those stupid cunts)

In fact, this kind of makes guessing obsolete, because it's pretty clear (b-but I guessed so far without reading through this, promise! I started with other posts!). This is not three-dimensional chess. This is a person who is smart – but not smart enough/too psychological damaged to be anything close to rational – writing a piece that is somewhat sophisticated and probably almost 100% authentic.

It is not true, though. Let's just get this out of the way before there are any misunderstandings, or before the lack of condescension in this post gives the wrong impression. This world view is not in fact true. R/C could have been raised in a lovely family and be naturally attractive and also lucky enough to have a functional relationship at 17 with a smart woman, and his views on gender would almost certainly be totally different. He wouldn't then necessarily be any more rational than he is now, he'd just have happened to have been born into a situation that more accurately reflects reality. After all, both worlds have identical truths about gender issues, but one yields to R/C objectifying women and the other to R/C respecting women. The problem is that R/C's beliefs are not causally determined by the state of the world, so much is up to chance.

This is, of course, a really fundamental problem. Most people's beliefs are heavily dependent on their immediate circumstances, and this often goes wrong because there is such high variance. Globally speaking, this often does yield to a majority of truth but... actually, let me explain this with an example.

Suppose a government passes legislation that affects all 10 000 people living in the country. Let us also suppose that the effects are net positive, and that everyone is affected equally, 6000 positively and 4000 negatively. Because humans are the way they are, the 6000 who are positively affected will at large support the legislation, and the 4000 who are negatively affected will, at large, not. So the change will have at least a 60% approval rating and it'll all work out. But that still leaves 40%, 4000 people, who are wrong, and even justifiably so in the sense that the legislation really did have a negative impact on their lives. And it also only works out if the set of people which are affected and the set which get to vote have sufficient overlap, and if the process is perfectly democratic. In real life, often enough, neither of those two things is true.

If you want your beliefs to reflect reality, head this way and read the sequences (or if you want to get to what is true directly, try this blog (or, if you want to skip ahead and just know what you should do to save the world, PM me.)) Otherwise, you just have to be lucky.

Fortunately, most people here will probably be fairly progressive when it comes to gender issues – by chance or otherwise – so I'll just take it for granted that woman and men are equal for all purposes discussed here and that sexism and objectification is wrong, and move on from there.

Okay, so R/C is being sexist on purpose, that much is clear, and this is exactly what gotta save women is about. In fact, the story is so over-the-top brazen about this that I have to give it a bit of credit – it is, in its own way, clever. And it is technically well done, both style and grammar are better than most other stories I reviewed in this group in the past.

The story is really two independent stories, both one chapter long. In the first, the hero gets sent to Equestria to save women from the devil. He meets with the main6, and there is a bunch of dialogue (which is, likewise, fairly competent). There is a jab at Spike, with Twilight telling him to stay behind, which at the time gave me a bit of hope that this could be subversive... MLP is arguably pretty bad about reverse-sexism in the early episodes, with the only two male characters being Big Mac (who can't talk) and Spike (who many think is a shitty character without a consistent personality). How cool would it be if the story made fun of MLP in that way by launching an overtly sexist premise that then gets turned on its head when the story enters Equestria, and suddenly Spike and the main character – the only males around – are the ones who are being objectified and need to be saved? That would be really clever.

Okay, let's get away from my imagination and back to reality. The devil plants a bunch of stinking mushrooms (there is something going on with mushrooms, they also make appearances in R/C's blog posts), rendering all the main 6 defenseless because their delicate lady-pony noses can't take the smell, and the hero beats the devil, whose corruption doesn't work on him because he – you guessed it – is not a women. And in the second story, the hero also gets sent to Equestria and kills Twilight and Fluttershy because he thinks they are his dolls animated by the devil. Again, though, I have to note how the dialogue and scene progression is competent. R/C really does stand above his work in that way, giving him some amount of justification to write it off as irony (even if the true agenda is clear). Me stating that it is sexist and offensive is clearly intended by him, and in this way I'm happy to satisfy that intention.

This is technically a review, so I have to give a score (but please treat this as as side-issue). Mind you that those scores are based solely on how much the story itself appealed to me – they are an observation, not a judgment. And the story is well written, has almost flawless grammar, and it made me laugh, so the score has to be positive, let's say 61/100. Yes this is the second-best score I've given so far, and I'm not happy about that either, but I don't blame myself. Whether we like it or not, this is also the second-best written story I've been reviewing so far. If you write something that is legitimately great, it will get a better score, even if that seems hard to do.

So there you have it. Raw Cringe, I can honestly say that I don't mind you that much more than most other people, though your worldview is not true and your existence probably has net negative impact on the world. Eh, actually let's not end it there, and I'll use the remaining space to talk about an interesting thought experiment that has absolutely zero connection to anything above: suppose a near-perfect forecaster came down from heaven and put two boxes in front of you, one of them transparent and the other opaque. In the transparent box you can see 1000$. The forecaster tells you that you have two options: you can take both boxes, or you can take just the opaque box – no-one is touching the boxes now, so there are no strings attached. However, before coming to you, the forecaster predicted what you would do, and she made her choice as to what to put in the opaque box based on her prediction: if she predicted you two-box, she has put 0$ in the opaque box, but if she predicted you one-box, she has put 1000 000$ in the opaque box. Now both boxes are filled, and your choice cannot change their contents. Do you take just the opaque box, or do you take both?

5877174 Any worldview based on a stereotype is inherently wrong, because all it takes is one exception to disprove it. It's really better to judge people on an individual basis. Most of the point Cringe made in his blog post can be argued down pretty easily, though I think on some things like the abortion issue he does have a point. One solution that's been proposed is that the father can have a "legal abortion" where he ceases contact and visitation with the child (as if it never existed), and in return he doesn't have to pay child support.

Anyway... mlp has a female cast, but tbh after the first season it really became a show for everybody rather than just girls. For most of the characters you could make them male and it wouldn't really be too different. Also I don't understand why cringe wanted a review from here so badly. There are tons of editors who will help improve a piece if you ask.

5877174 holy crap, that means this goes into the 'enjoyable' folder.

Woah ;p

5877280
And you win a million dollars!

5877470
That sounds accurate, it was enjoyable in its own way.

5877541 I'll have to read it sometime, when I'm not torturing myself with 2d animation

5877689

by the way mushroom = penis

... duh and it didn't even cross my mind.

and I would take the one box because I trust the forecaster

You also get a million dollars! But not from me

5877301
You're definitely going to find instances where women have unfair advantages, just like you're going to find people who have, say, 10 people they're close to, with a 50/50 split, and all 5 men are better people than all 5 women (in fact the odds of that happening are 0,12% so it has happened millions of times before). That goes back to the same problem I was talking about, you can get totally wrong conclusions by rationally observing your immediate environment just because of variance. I heard of the legal abortion idea too, it sounds fair (but I don't know much about it).

I'm not sure how much I agree with MLP becoming genderless in later episodes (though I actually like the feminine stuff). Stuff like Tanks for the Memories... I doubt this would work with a male cast. My impression was just that it gets a bit less childish. But I haven't gone over all episodes, so I'm not sure.

5877174 You should take both boxes, for several reasons.

First, the predictor is merely near-perfect. Well, so am I, on that particular prediction: if I predict that everyone will take both boxes, I'll probably get 99.9% accuracy. If I predict what people will do based merely on their opinion of Eliezer Yudkowsky and their familiarity with decision theory, I'd be even more accurate. The one-boxing argument requires perfect predictive accuracy--which is theoretically impossible to a Bayesian. A Bayesian can merely note the empirical performance of the predictor so far, which can be at most N out of N correct where N is some finite number. That converges on an accuracy of 1 as N approaches infinity--but is less than 1 for any finite N. (If you aren't convinced, imagine the predictor has predicted nothing, and you assign a prior of .5 to their getting the next prediction right. Update your prior every time they make a prediction. Your prior can never reach 1.)

Nozick says one-boxing doesn't require perfect predictive accuracy, but he says that by believing in abstract models more than in the laws of physics. If I make a prior for the impossibility of violating causation being right, based on all of the observations we've made so far on the subject, that's going to be a prior very near 1, and the predictor probably isn't going to have made enough predictions to counter it.

Second, you made no pre-commitment. Timeless decision theory is about making pre-commitments.

Third, the scenario posits a being which cannot exist--a perfect predictor. It is unwise to take scenarios believed to be theoretically impossible in our universe, and use them to try to prove things about our universe. Such scenarios have falsity baked into their presuppositions.

6244968
1. You walk out with a tenth as much money as I do.
2. the thought experiment is primarily relevant for two AIs that know each other’s code, and can therefore indeed predict each other’s behavior with perfect accuracy.
3.

if I predict that everyone will take both boxes, I'll probably get 99.9% accuracy.

Did you read previous responses?

6245364

the thought experiment is primarily relevant for two AIs that know each other’s code, and can therefore indeed predict each other’s behavior with perfect accuracy.

If they aren't deterministic, then they can't be predicted with perfect accuracy.
If they are deterministic, then the question is presupposing a contradiction: that they can, and cannot, make a choice.

Did you read previous responses?

People who read and reply to posts about Newcomb's problem comprise far less than 0.1% of the population. But I pulled that number out of thin air; my point was that the problem specification of a "near-perfect" predictor was too vague to justify your claims. 90% accuracy would probably be called "near-perfect" in prediction of political races.

6245717

If they are deterministic, then the question is presupposing a contradiction: that they can, and cannot, make a choice.

It doesn’t matter whether they make a choice. What matters is whether two AIs will cooperate in a one-shot prisoner’s dilemma.

Two AIs with functional decision theory will. Two AIs with causal decision theory won’t. (Given they know each other’s code).

6246465

It doesn’t matter whether they make a choice. What matters is whether two AIs will cooperate in a one-shot prisoner’s dilemma.

That's not the question you asked, nor the question you answered by saying "You walk out with a tenth as much money as I do." The case of two deterministic AIs is not comparable to the problem for humans. Newcomb's problem, and a wide set of other problems, when applied to humans, are about the difficulty of using pre-commitment. There is much less difficulty when you're programming deterministic AIs who can read each other's code. One-boxing may be a good solution, although the claim that you can "just simulate" what another AI will do in the future by running its code in a sandbox now is naive and wrong, because you can't know what the future context will be.

But in any case, while this consideration of deterministic AIs is interesting and important, it does not imply that humans "should" one-box.

6247965

although the claim that you can "just simulate" what another AI will do in the future by running its code in a sandbox now is naive and wrong, because you can't know what the future context will be.

You’ll know what the context is once you’re in the situation. If you then also know the AI’s code, you should know enough.

But in any case, while this consideration of deterministic AIs is interesting and important, it does not imply that humans "should" one-box.

Yeah, true.

Honestly though, you don’t need a perfect forecaster to make one-boxing preferable. If your forecaster just has an 80% chance to successfully predict your algorithm, one-boxing already beats two-boxing. The only loophole is that now you could claim to be among the 20% and therefore 2 box and get the bigger prize, but unless you claim that your algorithm has zero baring on what they will conclude trying to guess your algorithm, you can always tweak the problem such that a one-boxing algorithm has better expected utility.

6248018

You’ll know what the context is once you’re in the situation. If you then also know the AI’s code, you should know enough.

You don't get to simulate when you're at the decision point. You have to simulate at the point of pre-commitment.
But in fact it doesn't matter when you get to simulate, because the simulation takes time, and you have to decide whether to trust the AI after you finish the simulation, by which time you're in a different context and your simulation result is no longer a proof. It is strong evidence, but the discussion on LW pretends it's a proof.

Honestly though, you don’t need a perfect forecaster to make one-boxing preferable. If your forecaster just has an 80% chance to successfully predict your algorithm, one-boxing already beats two-boxing.

Preferable as a policy, not as a decision-theoretic claim. Decision theory is a precise model which states what is the best decision at a moment in time. The question being asked is what is the best decision according to decision theory, not what is the best policy to take at some prior point in time. That may not be the most-important question to ask, but it is the question implied, by default, by our default model of decision theory, if someone asks "Should you one-box?" without fully specifying the problem.

An example of where the distinction is important is if you aren't told the rules of the game until after Omega has made its prediction. In that case, two-boxing is the best decision.

6248329

because the simulation takes time,

Yeah... like, 0,001s for an AI?

Preferable as a policy, not as a decision-theoretic claim. Decision theory is a precise model which states what is the best decision at a moment in time. The question being asked is what is the best decision according to decision theory, not what is the best policy to take at some prior point in time. That may not be the most-important question to ask, but it is the question implied, by default, by our default model of decision theory, if someone asks "Should you one-box?" without fully specifying the problem.

Aren’t you literally just defining a decision problem so that causal decision theory will win out every time? (Even though it systematically performs worse or equal than FDT at every issue.)

But sure, by that definition, two-boxing might be the “better decision”, but one-boxing is the algorithm that does better. One-boxers do better than two-boxers, and everyone has a chance to decide what she wants to be.

6248405
Eliezer's argument depends on being able to prove that you know what an AI will do. 0.001 second is enough time for something to change, and any change makes the simulation results no longer a proof. The entire point of invoking determinism is to move from mere evidence to proof, but you can't.

But sure, by that definition, two-boxing might be the “better decision”, but one-boxing is the algorithm that does better.

You and everybody on LW keeps saying that. What I have been arguing here is that that is an incomplete specification. One-boxing is a winning algorithm under some ways of implementing the game, but not under others. Explained above.