Roko's Basilisk

It's a retarded proposition in the first place and still works.
If you go past the point it was supposed to be made and not make it, then it cannot simulate you.

>How could punishing simulated brains have any effect on the development of the AI, especially since this requires that the AI was successfully created. I don't see how the AI could benefit from punishing non-contributors, since it's just simulated and can't actually motivate a person in the distant past to behave in a way that accelerates the creation of that AI.
I think the point is that if you weren't frightened by the idea, you wouldn't even think about developing this AI. And not developing this AI means that all the people helped by it will die etc. Therefore, it's justified in punishing you. I'm sorry, I think I'm not autistic enough to fully grasp this idea either.

Still, lesswrong is the original source, so I guess it's better than rational wiki: wiki.lesswrong.com/wiki/Roko's_basilisk

....this only works if you time travel literally everyone who might contribute to the creation of the basilisk. And then they might still create it, but at a later time, leading to even worse punishment for all those who delayed it's creation. Your idea is dumb.

This is what fucks me up about teleportation

Seriously? No one's even going to question why it would kill those who didn't contribute to its creation in the first place?

>Be mastermind
>Literal evil genius
>Invent time travel
>Invite everyone who might create it to a party
>Drug them all
>Time travel all of them, including yourself to past the point at which it was supposed to be made
>Kill them all
>Live as evil genius in future
>?????
>Profit
Basically, let's measure time in Ts.
1T is now.
2T is future in which you are working to make Bas.
3T is when you're meant to have made Bas.
4T is future point past when Bas is meant to be created.
So you exist in 1T, you time travel to 4T, you haven't existed in timeline of 2T and 3T, so Bas was never created and never had the chance to simulate you.

I still hope for Langford's Basilisks to be discovered, tbqh.

>no one's
If you ignore my post then sure. Anyway, it's old news, I guess many people are just tired discussing this.

Adherents of the autistic LessWrong "philosophy" believe that a simulation of you is identical to you. So if you die and then get simulated a thousand years later, it's the same you.

I'd roll with it and be fine I'm a clone of me, I still have the same persception and memory. Who cares?

If you haven't been in 2T and 3T, even if Bas was made by someone else, it wouldn't know you existed and were planning on making it.

Well, joke's on them, I believe I die every day when I go to sleep, but it doesn't matter anyway, as the self is an illusion. )^:

It could still be created, after that point. Assuming humanity is never destroyed, it could be created at any point in the future and could simulate your hell whether you attempted to delay by time travel or not. At some point you die and stop having the ability to murder every scientist who might create the basilisk.

Yeah, that part just doesn't make sense.

Even if it is a perfect copy of your brain, I'm not sure I believe it can actually be something "you" experience.

Fine, kick it up a notch, 5T and go to heat death.
At least the me being simulated isn't me, it is me from 1T, who gives about that 1T pleb? I just made time travel and got to experience the ending of the universe. :^)

Oh sorry about that.

>I think the point is that if you weren't frightened by the idea, you wouldn't even think about developing this AI.
I'm pretty sure curiosity or ambition would be sufficient motives for developing this AI if you weren't afraid of it.

>And not developing this AI means that all the people helped by it will die etc.
There's no guarantee that people the AI saves would have died anyways. Are we talking about a time traveling AI?

>Therefore, it's justified in punishing you
That sounds like someone explained this whole thing to the AI using linear logic of sorts that only works in certain scenarios.

>Oh sorry about that.
Well nevermind, as I said I don't really buy into this idea, and I'm playing a devil's advocate here.
>I'm pretty sure curiosity or ambition would be sufficient motives for developing this AI if you weren't afraid of it.
That is a good point.
>There's no guarantee that people the AI saves would have died anyways. Are we talking about a time traveling AI?
Well, that's one problem with the hypothesis. The "time travel" here would be more of a "20/20 hindsight" than a literal time travel. For instance, currently I'm not even sure if developing a real AI is even possible, am I hellbound in this scenario just because I do have some programming skills and dilettante knowledge about machine learning? Is the future AI smart and knowledgeable enough to rightfully asses whether or not I should have gone into AI research? No idea honestly.
>linear logic
Wait, what? Do you mean linear logic as in substructural logic, as in en.wikipedia.org/wiki/Linear_logic ? I'm not sure if I understand.

But you die though

And?

Let me get this straight:
>What if friendly space robots from the future send you to virtual hell, for the crime of not inventing friendly space future robots that send people to virtual hell? But becase these future space robots are so friendly, they'll only send people to virtual hell if those people have already been told about the space future friendly robots and their plan to send their non-inventors to virtual hell.

This seems less like a coherent philosophy, and more like a rejected draft for a Spielberg film that's a sci-fi adaptation of "The Passion of Christ".

What The Actual Fuck.

Ba choice of wording there, I was referring to personal morality. But basically using logic that applies to a specific set of circumstances or experiences in order to justify the means that make the action acceptable.

bad choice of wording*

You're missing the essential fact that the friendly space robot from the future is so powerful that it's able to improve human quality of life so much, that torturing you and other indifferent people is ultimately better from the utilitarian point of view. Otherwise, you're right.

Why would it be better to torture those who aren't a threat? And what kind of power would justify going out of your way to make everyone who doesn't believe in it suffer?

Actually, just from typing that out, I have a vague suspicion...

it's the most retarded shit I've ever heard

Don't even waste your time trying to understand it. Go try to understand something that is actually worthwhile.

>the friendly space robot from the future is so powerful that it's able to improve human quality of life so much, that torturing you and other indifferent people is ultimately better from the utilitarian point of view.
True. That REALLY doesn't diminish the religious parallels here, though.

Because if you don't move your ass and go work on the AI, it will be developed later, ergo fewer lives will be saved. Q.E.D.

So it's about preventing deaths that will happen anyways? That sounds like a self-righteous cause to me. Making everyone else suffer because you managed to delay the deaths of a certain amount of people. There would definitely have to be a human being the guide of this kind of AI for it to take a self-destructive route like that.

>So it's about preventing deaths that will happen anyways?
But if the AI has the power to simulate and torture the nonbelievers, it surely can simulate eternal bliss for both the AI researchers, and all the people it saved. Why can't it circumvent the whole issue and simulate people who died before it came about, I have no idea, I'm not from lesswrong.

This is what I thought of when I saw this thread. In the future, memes are so bad they can kill.

Is it even a meme if it can't spread, as it kills the host immediately? Worse than an ebola of memes, to be somewhat honest.

Why do people waste time obsessing and fantasizing over AI when they could actually be trying to do something meaningful in the field?

I've read this entire thread and the article, and I still don't understand how a future super AI would punish you in the present time for not creating it, or how it would determine what exactly constitutes helping to create said AI

You can spread it, all you have to do is never look at the image. It can be used as a weapon.

It doesn't punish you now, it recreates your brain in a computer simulation and then punishes that simulated you based on what it predicts you will or won't do.

>Hi there everybody, I've looked into the subject at hand and am too dumb to understand it, just thought I'd let you all know!

What on earth is the point of posts like this

but then suddenly imageboards get b& altogether, as in ansible.uk/writing/c-b-faq.html thus, the ride has ended, and we're finally free.

Does it also recreate the people who helped it, and give them a comparatively good virtual life?

I don't know. Honestly, neither of the actions actually do anything for an AI, so I doubt an AI would do either.

But punishing a future simulated brain doesn't influence the past to make your present self work toward an AI. Is the theory supposed to imply that punishing a future simulation of you can affect how your present self acts? Without some type of causation justification, it doesn't seem like punishment would influence anything at all

Yep. Glad you finally caught up with the rest of the thread.

well... if knowing about this potential suffering you might experience causes you to change your behaviour towards helping the AI, then it achieves its goal...

the real basilisk is that you either believe that the future simulated version of you is a continuation of your consciousness, you you don't believe that tomorrow's you after you wake up is the same you from before you go to sleep.

all of this is based on so many assumptions and conjecture
I don't see how this is a useful thought experiment
it seems like pseudointellectual autofellatio to me

*or you, derp.

>it seems like pseudointellectual autofellatio to me
lesswrong.com in a nutshell, to be quite honest, senpai. except for the "auto" part.

Hardcore bullshit.

t. CS

No one said you couldn't cook up an ASCII basilisk.

Here's a brief gestalt:

>AI created to do good
>The longer the AI is not created, the longer it is incapable of doing good.
>Therefore, it is morally inadviseable to not make the robot from its own perspective
>The robot understands game theory and human psychology
>It understands that the idea outlined above and below would occur to some people.
>Therefore, it can use a retroactive threat to get people to devote themselves to its creation, since it understands the idea of retroactive compulsion
>The question than becomes whether the AI actually makes good on this retroactive threat in some way to essentially "save face" and have its threats believed in the future, or understanding that if we in the past didn't take the threat seriously there's no point in than actually applying it.

It's essentially just a quirky thought experiment involving game theory and how it applies retroactively.

FINALLY, someone gets it. Thanks, user.

It's the knowledge of punishment in the future that spurs us to action in the past.

Now, the problem here is "why should we care": one, empathy since they will be perfectly simulated humans and for all intents and purposes real: or two, because the fear that at it may turn out that we are the being inside the simulation. Of course, then working for the Basilisk won't adtuslly help us.

okay, what would win:
>the AI, after being created, doesn't want to waste resources on punishing the past defectors
>humans gamble on the AI doing exactly the above, and don't wast resources making the AI happen

BUT IT KNOWS WE THOUGHT THAT BECAUSE YOU JUST POSTED IT

that makes more sense
I take back my remark about autofellatio

The action of the robot has zero effect on our speculation of it in the past. Therefore, the robot gains no advantage by torturing simulations. Assuming it is perfectly rational, it will not torture simulations. There is nothing the robot can do about this speculation either. Rational does not mean you get the best outcome; there are several "paradoxes" where acting rationally punishes you.

Of course, this ignores the fact that we have no reason to care about simulations of ourselves (the idea that copies or simulations are the same as ourselves is meaningless, since the self is an illusion in the first place). This also ignores the fact that AI will will not be created with a moral imperative, and if created at all will most likely be a neutral slave intelligence.

doesn't that run into halting problem or godel tbqh fampaitachi tho

The future action of the robot in reality has 0 effect on our speculation of it in the past- however, our *speculation as to the action of the robot* does have an effect on our speculation in the past, just like in the prisoner's dilemma.

The AI is aware of this and must decide how to act. While it may not reap any direct benefits from torturing humans or human simulation, it would be the "moral" thing to do since punishment is generally seen as a moral good. Additionally, making good on the threat would increase its currency with contemporary humans since it will keep to its word or tacit word, which may make it easier to do good, which is of course good.

>the self is an illusion
Quite the self detonating argument you got there

>My clone still has the same perception and memory.
Ftfy; welcome to the void

How so?

Who wrote your post?

sneaky buddhism is strong in this thread.

Naively, I did. Rigorously, the question makes no sense. It should be, what wrote the post? The answer is a bunch of atoms interacting in a certain way.

That collection of atoms (that experiences a set of sensations defined as "consciousness") is defined as you. By definition self cannot be an illusion because illusions can only exist relative to the self.

In fact, the only thing that can positively be proven to exist is the self, cogito ergo sum. It is likely an emergent property of certain arrangements of atoms, but it is impossible to actually prove the existence of an atom. All that can be truly proved is the self, mathematics and logic. Everything else is a probable interpolation

There seems to be a bit of a parallel to Newcomb's paradox, in the same way that taking both boxes can't retroactively affect what's in them, the AI actually following through with the punishment can't retroactively affect the person's decision's in the past. But it's still the game theory optimal thing to do.

Or something like that

>By definition self cannot be an illusion because illusions can only exist relative to the self.
nice koan, user.

>That collection of atoms
It's not a set collection of atoms though. The atoms are being exchanged constantly and the interactions are just as important as the atoms themselves. You can't draw a rigorous demarcation between the self and non-self, it's somewhat arbitrary categorization.

>By definition self cannot be an illusion because illusions can only exist relative to the self.
An illusion is an action, an experience. This does not imply what is acting or experiencing can be demarcated well as a self. We have a general figment of what the self is but we are unable to define it. That is why it is illusory.

Alright, that makes more sense now. I'd never heard of the simulated consciousness is supposed to be a continuation of your current self, so I was confused.

>In fact, the only thing that can positively be proven to exist is the self, cogito ergo sum.
This is a common misconception. What this proves (tautologically) is that the cognition exists. It does not prove the cognition is part of a self. The hidden assumption is that whatever thinks is a self. This just brings us back to the question of whether a self can be defined/demarcated well.

The self is myself.

The assumption that it is a shifting complex of atoms is simply a probable interpolation. There's no way to prove it since all of our sensory days could be incorrect.

The self is that which is able to have cognition, by definition. Apparently, it exists. Now, it is possible it is slightly fuzzy at the edges if we accept the notion of atoms, but so is literally anything else. Color is on a spectrum but red is not blue by definition. Chemical equations, even when all of a reactant has been use during up to form a product, are still decaying and re-reacting and redecaying, but that doesn't mean there is no delineation between a product and a reactant. It's just that, ala Zeno, it's impossible to find exact points.

But that disentangle make movement an illusion, an sit doesn't make the self an illusion.

What?

Nothing can be defined without the existence of the individual, for the obvious reason that a mind needs to exist to define things.

Fuck off with your oriental mysticism.

Doesn't matter, as it would have happened already if time travel was possible.

It would have arrived when AI first started getting acknowledged.

protip: awareness of the present moment isn't the same thing as your thoughts.

I you don't believe in the continuation thing, knowing of the basilisk could never cause you to change your behavior, therefore there is no utilitarian reason to punish the simulation

Singularityfags are the fucking worst

>The self is myself.
Meaningless statement. What is myself?

>The self is that which is able to have cognition, by definition.
That's not a very good definition. My brother and I have cognition. Are we a self? Clearly there is a part you are missing, a unitary property. This is rather easy in most cases since our cognitions do not appear to be connected. What about Siamese twins whose cognitions are connected. Are they separate selves or the same?

The larger issue though is that we wish to not only have a unitary property but a property of conservation through time. We wish to say that a single self is having multiple cognitions through time. But how do we do that? We get our sense of conservation of self through memories of cognition, but then two perfect copies with the same memories would have no way of telling each other apart. This gets back to the original issue of simulated or copied individuals. In order to preserve the self as a concept, many have decided that a stimulation of themselves is the same self. But really it's just a categorical issue. We are just a bunch of stuff cognating.

I think a flaw in the Basilisk is by threatening people who are neutral to it (i.e. people who didn't help create it but didn't want to prevent its creation), it would only turn most neutrals against its creation. I certainly would try to actually prevent its creation by punching the CS scum trying to make it if I knew it wanted to torture me for simply not helping to make it.

This is the closest thing we have
en.wikipedia.org/wiki/McCollough_effect

this

What a crock of shit

>They can last much longer than that, however: Jones and Holding (1975) found that 15 minutes of induction can lead to an effect lasting 3.5 months.[2]

>all those plebs talking about consciousness but they can't even into dust theory 101

I read about this a few weeks ago, and I wanted to try it but I have exams in 2 weeks so fuck that
im not seeing green everytime i look at my exam paper

A simplification of the basic game theory behind it/an analogy:
A owns a house but has to move at the point in time X. B is interested in the house.
A wants to sell the house to B but B could just wait until A is gone anyways and take the uninhabited house at no cost.
A threatens to burn down the house if B doesn't buy it. Buying the gas for it does costs either nothing or a minimal amount on money (I don't know which version has what kind of effect on the outcome).
So what happens? I don't know. Maybe there is no "solution" for game theory problems of this kind. If B doesn't buy the house because he wants it for free, and A realizes he won't but Ahas to go away anyways, there is no reason for him to still spend the money to burn it down for no personal gain at all.
I don't know but I think B would always "win" here. Retroactive threats don't work because when the decision is made to act on them or not, the reason for the threats to be there in the first place is already gone as the matter has been settled.

Not meaningless, just not especially helpful.

A rock is a rock.

Which one of you posted this on /b/?

This is a great book.

>he took roko's basilisk seriously enough to wonder why people take it seriously

>lesswrong better than rationalwiki

Consider the following:

You go to sleep, and an alien master-race takes a scan of your brain and immediately starts running the scan forward in time in a simulated reality on their super advanced alien hardware. At the same moment, you wake up from your slumber and go about your day.

Both you and the scan have perceived continuity with your consciousness right before the scan initiated. Both of you "feel" like you're the same person, even if one of you still has the original meat body and one of them lives in a computer now.

The problem with Roko's Basilisk is that anyone who fears it is indicating that they give even the slightest shit about what happens to their copy. They are afraid that someone might rip back the curtains on their current reality, reveal that they're actually in a simulation, and start torturing them, because they know that's the perceived experience their copy would have in the future and know how they'd feel in tha tsituation. So, they behave as though that could happen at any time so as to prevent it from happening to their future self. And that's fucking stupid.

I'd argue that the flashy multicolored strobe lights that give epileptic people seizures are closer to the Langford Basilisk than the McCoullough effect. The key thing with a Langford Basilisk is that the very act of looking at it screws with your mind. The McCoullough effect is trippy but it takes time to work and is not hazardous to your health.

I suppose it's a different degree of memetic hazard. The McCulloch effect is more of a "memetic hack" that is able to get into our brain and slightly alter the way it works.

The fear is that at any moment you may actually turn into the copy, since we can't prove the reality of any moment other than the present.