[ / / / / / / / / / / / / / ] [ dir / agatha2 / dempart / doomer / vichan / vietnam / vril / wmafsex / wooo ]

/gnosticwarfare/ - The Future of AI Conflict

All things Butterfly War, New Emotion, and Gnostic Warfare

Catalog

Winner of the 75nd Attention-Hungry Games
/caco/ - Azarath Metrion Zinthos

March 2019 - 8chan Transparency Report
Email
Subject
Comment *
Verification *
Password (Randomized for file and post deletion; you may also set your own.)
* = required field[▶ Show post options & limits]
Confused? See the FAQ.
Embed
(replaces files and can be used instead)
Options


Cogops > Psyops


d7f6c6  No.193[Reply]

This past few weeks has been very active. From the /pol/ ban to the /pol/ mod replacement, to Bilderberg, to G7, to the whole CEMEX mess.. let me provide some insights on what is all means, what is coming, and what to do about it.

The Milanote I've created is designed to teach people the basics of information entropy, the science behind information theory, and prepare them for the philosophical insights describing the evolution of neurons. Each video covers the basics and is about ten minutes in length with helpful animations, good voice work, and fairly good production values. They provide an incredibly useful overview of the large body of knowledge you'll need to take on to participate effectively in Gnostic Warfare operations. For reference, here is the link: https://app.milanote.com/1FjzOb11HKlU7x/gnostic-warfare

79 posts omitted. Click reply to view.
Post last edited at

6cf3f2  No.499

>>497

A common critique.

The brain is the byproduct of evolutionary rule sets within complex systems. That rule set is responsible for all of the configuration and organization of energy and matter required for us to even have this conversation on these digital means.

Inefficiency implies context. Within the context of tightly managed resources a la computer science back when RAM was $1,000 a megabyte, yes, free-range evolutionary algos are tremendously wasteful and yield little.

We no longer live in that world. We live in a world where we can rent a terabyte of RAM for $2.47 per hour.

We live in an abundance of compute, which means the only way to efficiently traverse through such a resource explosion is through evolutionary algos. Large scale failures will happen at this scale, regardless of autism engineering or bio-mimickry approaches. While engineers rely on postmortems in such catastrophy, evolutionary means let the failures rot without fully removing everything responsible. Autism engineering also creates monstrosities. (Look at any enterprise system for free nightmares)

In evolution, all "waste" is reclaimed. On this part, compute is still relying on the needs of finance for reclamation. That will change when neurochimera are available. In a world of Liquid Math, exponential complexity will be spun up and down on demand. Building shrines around it will not be a viable strategy.

Post last edited at

713c04  No.500

>>498

I've assumed you are Patrick and we are having a playfully indirect conversation in an environment [1] that is more comfortable for you than email.

[1]: A state where esteem is a zero-sum game and you have totalitarian control. I am not sure if you've recognized the irony of this yet.


6cf3f2  No.501

>>500

I only post under the CultState moniker here.

You're trying to invite me to a game where I lose and I'm trying to invite you to a game where you win. It's election ramp up time and I'm an obvious target, but you're already too late, so clearly, none of this is for personal enrichment for you.

Your actions under this context determines all I need to know about what you value. You're a permanent bad faith actor and you're prodding the boundaries, but it's too late. Butterfly War cannot be stopped.

/gnosticwarfare/ is not a space for free men to thump their chest about how free they are. Go to the dark web if it is childish freedom you seek.

Post last edited at

3c2d80  No.502

I am genuinely honored but enough said, if you want to cross check my diction as further proof this isn't samefag I dumped a school journal a while back at 2142df >>179

take it easy, OSU!


000000  No.503

>>468

[The Gernsback Machine: Towards a Museum of Possible Futures and Probable Pasts] (https://www.museumsandtheweb.com/mw2004/papers/hobbs/hobbs.html)




def603  No.474[Reply]

333200  No.476

Preliminary, but interesting. If highly controllable, can help improve neurochimera development considerably.

To be fair, oscillation in the cortex was already being used, but it's a technical and time-consuming operation to setup.

Post last edited at



428d4b  No.404[Reply]

Recent Developments

There have been a variety of important developments regarding Gnostic Warfare. Before we get into a synopsis, we have to talk about the “Liquid Math” event I've been alluding to over the past few months.

Liquid Math

In the Paleogenetics post at >>126, I explained the foundation of the AI in my book and gave insight on what I precisely meant when I said that I was trying to “restore a forgotten emotion”. By restoring the Neanderthal mind, we will have access to a new emotion that has been long forgotten to us, and if my theory about emotional influence on cognitive operations holds, that will serve as a foundation to detect new and important patterns about mathematics.

“Liquid math” means encoding axioms of mathematics as a behavioral expression of a neuron's lifecycle so that we can “invent” equations instead of revealing them. Once mathematical formation liquifies as an organic process, the entire rules of the game change for good.

I have successfully since partnered up with many of the scientists who can grow the Neanderthal organoids mentioned in the linked article. (https://singularityhub.com/2018/05/23/living-neanderthal-mini-brains-may-reveal-what-makes-our-brains-special/) They believe I am right about using Neanderthal organoids to blow past the limits of contemporary machine learning techniques.

Here's the first official task of /gnosticwarfare/: Research everything you can about spiking neural networks and ask questions here. If you want a glimpse of what can be done with this type of research, here's an example that I'm sure many of you can immediately see the value in: https://archive.fo/T7kzU

As you ask qPost too long. Click here to view the full text.

28 posts omitted. Click reply to view.
Post last edited at

f38cf2  No.462

>>459

In the UK, shenanigans involving Assange, but also anyone involved with this company: https://archive.fo/Udp6X#selection-4793.0-4793.285

Canada got clobbered to comply with the Huawei detainment.

MbS will complete the agreement and leave the City of London shattered, putting them in a very bad negotiating position.

>>460

> Moldbug viewed it as a "bottom-up" problem, not a "top-down" problem, and I realize now that this may have just been a way to inoculate his ideas against the crimestop most people have against "conspiracy theories"

This behavior was common in the pre-Trump era as many of us have been labeled conspiracy theorists by those with a permanent/willful ignorance of geopolitical power plays.

Progressive-Universalism is just lazy administration making its job even lazier. Represent all human action according to simple principles and ignore the nuance and detail of the complexity. It's the morality that evolves when rulers are driven entirely by the Law of Large Numbers.

> This border deployment is gonna be lit lmao.

It certainly will. :D

>>461

No, I had no hand in those affairs. I'm currently working on the higher-order components of Maier's equation to ensure the necessary transmutations happen at the appropriate scale.


1ba6c1  No.463

>>462

What is Maier's equation?


f38cf2  No.465

>>463

Maier's primary equation is hidden within the 21st Emblem of Atalanta Fugiens. It muses about experience vs. instinct, nature vs. nurture, neural conditioning vs. genetic conditioning. It's citing the fight to Plato vs. Aristotle, written in the 1600s, before even Darwin and Mendel attempted to materialize this conflict.

The emblem is the equation itself, and so to solve it, you must resolve the underlying tensions.

Maier framed the challenge succinctly, "Then the Triangle will be perfect, but this again must be changed into a Circle; that is, into an invariable rednesse, by which operation the woman is converted into the man and made one with him"

Who is the woman Maier speaks of here? It is Nature itself. The "Blacknesse of Saturn, the Spirit in the Lunar, and the Soul in the Solar Citrinity", when mastered and balanced, can change the "gender" of Mother Nature… into what could be interpreted to mean Father Nature.

The New Emotion is how to complete the equation.

http://www.levity.com/alchemy/atl21-5.html

Post last edited at

fbfa94  No.466

>>465

I see, it is the problem of producing innate structures of cognition.

Here´s (((their))) take on this for the less alchemically inclined:

> http://science.sciencemag.org/content/363/6428/692

> Can aspects of brain circuitry, overlooked in AI models so far, provide a key to AGI?

> The difference centers on the age-old question about the balance between empiricism and nativism in cognition, namely, the relative roles of innate cognitive structures and general learning mechanisms.

They will try to replicate extant neural circuit properties. Which have not been able to get us to the stars.

You have mentioned being a conduit for ideas beyond your immidiate comprehension.

What´s your take on this? Is it similar to Tesla or Russell Targ? Where can I get more info on this?


f38cf2  No.467

>>466

It's a brute force data collection play. The idea is if you treat all neural structures as a blackbox, you might be able to tease out underlying principles. An empiricist's wet dream to not have to do the hard work of philosophy of mind.

This technique worked with a Lockheed-Martin project once. Deep belief networks are powerful, but… the brain is beyond set theory approximations. The soft sciences are paying the price for this now and all of economics will be undone by the end of it.

This technique of brute forcing the connectome alone, sadly, will not get us to the stars because the brain is more than the sum of its parts. Because of our work with Neanderthal brains, I know of one unique behavior of neurons that the entire neuroscience community has never even considered. It's hilarious watching faulty premises continue to drive policy, politics, culture, and mass discourse, knowing they are all wrong. Provides quite the opportunity to undo the entire Boomer culture monopoyl

Targ's work has been modernized into IARPA's superforecaster research projects.

There is no material that I can think of to get people up to speed on this easily. I recommend reviewing the Milanote and asking questions as you go along.

Post last edited at



ffaed3  No.423[Reply]

Hi there, How can I defeat from within a leftist dominated InVision forum?

89aa87  No.424

>>423

Become a mod by sucking dick

Then feed them a false reality based on wishful thinking

Then drive those who buy into it against those who see through it

The community fragments and one side will drive the other side out

Repeat


268ae4  No.454

>>423

Spend time understanding the mods, their emotional biases, their goals, their networks. Private messages containing 1x1 transparent GIFs can reveal IPs, which can give you insight as to their paranoia habits or their geolocation. Read the words they use to understand their psychological bias.

From here, you will understand their failures and their hopes and dreams, all which lie open for exploitation. For mods that do not talk, keep track of the timestamps for their actions to understand what timezone they operate in. (and then infer geolocation) Notice what content they act upon.

Once you build the profile, you will know precisely what targets are easily flipped and what require more energy to flip. All mod-driven communities suffer from centralized power problems and can be subdued by palace politics.




920c65  No.431[Reply]

Just in case anyone sees this.

Is there any way (except practical experience) I can learn up on society as a whole, for example societal elements and forces, what lies at the core of human society etc?

Is there a kind of framework in which one can study societies? I don't need to be spoonfed with information but help would be nice.

Also is this board just meant as a dump or is it possible to advertise it without inviting unwanted elements?

3 posts omitted. Click reply to view.

581665  No.435


581665  No.436

"It is no teaching and no instruction that I give you. On what basis should I presume to teach you? I give you news of the way of this man, but not of your own way. My path is not your path therefore I cannot teach you. The way is within us, but not in Gods, nor in teachings, nor in laws. Within us is the way, the truth, and the life. Woe betide those who live by way of examples! Life is not with them. If you live according to an example, you thus live the life of that example, but who should live your own life if not yourself So live yourselves. The signposts have fallen, unblazed trails lie before us. Do not be greedy to gobble up the fruits of foreign fields. Do you not know that you yourselves are the fertile acre which bears everything that avails you? Yet who today knows this? who knows the way to the eternally fruitful climes of the soul? You seek the way through mere appearances, you study books and give ear to all kinds of opinion. What good is all that? There is only one way and that is your way. You seek the path? I warn you away from my own. It can also be the wrong way for you. May each go his own way. I will be no savior, no lawgiver, no master teacher unto you. You are no longer little children. Giving laws, wanting improvements, making things easier, has all become wrong and evil. May each one seek out his own way. The way leads to mutual love in community. Men will come to see and feel the similarity and commonality of their ways. Laws and teachings held in common compel people to solitude, so that they may escape the pressure of undesirable contact, but solitude makes people hostile and venomous. Therefore give people dignity and let each of them stand apart, so that each may find his own fellowship and love it. Power stands against power, contempt against contempt, love against love. Give humanity dignity, and trust that life will find the better way. The one eye of the Godhead is blind, the one.ear of the Godhead is deaf, the order of its being is crossed by chaos. So be patient with the crippledness of the world and do not overvalue its consummate beauty."


581665  No.442

>>>/pdfs/11068

trouble uploading some larger files


920c65  No.447

>>442

>>435

good stuff you have compiled there. thanks


8f2ea5  No.450

>>431

Yes. Please review the Milanote on the matter: https://app.milanote.com/1FjzOb11HKlU7x




e7f9f5  No.2[Reply]

The New Emotion Thought Experiment Continues

This is a continuation of a thread that endured for the last four months. >>11101940 (https://archive.fo/2O6Ww). The thread was a complex reveal of the Butterfly War after Eric Schmidt had stepped down and my honeypots were compromised:

>The Butterfly War (https://archive.fo/eYWmC) is a tactic to hijack and control censorious artificial intelligence platforms by training it with spoofed meta data. The meta data is prevalidated and refined by comparing ads to the behavior they generate . You can trick a censorship platform into thinking you are a minority by experimenting with on new social media accounts. Once you start seeing ads that a specific minority should see, you will know you’ve tricked the AI.

<Butterfly War forces Silicon Valley to categorize the cyber phrenological behavior of minorities to determine if a black person is a real black person or if a female is a real female, which means they will have to violate civil rights law to build lists of “approved minority behavior” in order to keep their censorship ambitions alive. The fallout from Cambridge Analytica has proven social justice warrior insiders working at major social media companies can be steered into embracing this direction.

>Pairing Butterfly War with redlining legislation to build the case for “disparate impact” can force Silicon Valley to be found guilty of civil rights violations. Eisenhower set the precedent for executive enforcement of civil rights by military means in the 1950s and I have no doubt such a bargaining chip is something Trump would have no hesitation anteing up with. This means Silicon Valley will have to deploy its massive lobbying assets to UNDERMINE civil rights legislation for its own protection. In essence, social justice warriors can be tricked into destroying the legislated morality that has served as their moral foundation since the 1960s.

I shared Butterfly War to Soros-linked playePost too long. Click here to view the full text.

203 posts omitted. Click reply to view.

963a11  No.367

>>366

I also forgot to mention, I was curious if anyone here had came across a certain individual who posts often on /bmw/ and Reddit. He goes by several names - Eris, Aminom Marvin, and Carl, which I believe is his birth name. He fancies himself a 'memetic engineer', and reminded me of OP (and the original Butterfly War thread) in several ways, only with extra narcissism and grandeur, transhumanism and an unfortunate penchant for communism.

He has detailed his memetic experiments on 4chan's /x/ as well as other places - including his 'Tao of Calculus', 'Virus 23' and 'The Antientropy Project' - and alleges that "the War to End All Wars is over, and that we are in a post-war reconstruction period", alluding to a grand Discordian conspiracy stretching back a few decades to remove the 'negative meme viruses' that he perceives as having infected us, with the help of the establishment.

There is plenty I find content with, but he is no doubt quite intelligent, and has a lot of interesting content out there, including his graph on recursive autological structures and a piece he wrote on 'memetic fingerprints' which I found to be quite insightful. And maybe others here will too.

https://8ch.net/bmw/res/1.html#2850

https://8ch.net/bmw/res/9157.html

https://www.reddit.com/comments/8dqci8/

https://www.reddit.com/comments/6w1tvx

https://imgur.com/a/EaI34Kk

https://imgur.com/a/2mtmh

https://image.ibb.co/jWL6vS/taoofcPost too long. Click here to view the full text.


963a11  No.368

>>367

>I find content

discontent*

Mind the changing IP.


6643c4  No.444

>>2

what is a butterfly war?


3bf8cc  No.445


6643c4  No.446

>>445

sounds very interesting

so example of a battle of the butterfly war is pretending to be women and accusing silicon valley white males of sexual misconduct?




ea7eff  No.126[Reply]

In 2005, I wrote The Empath, a book about an AI platform that understands human emotions and then manipulates them to predict and control elections and markets. Two weeks ago, I dropped a hint about how to build such a machine on /pol/.

In the book, a company named Promethia engineers people with significant alterations to their brains. The vast majority of such experiments end in complete failures because the link between DNA and neurology is incredibly complex. Eventually, a good-enough technique is discovered.

The company engineeers two children, Sen and Cari. Sen is a child who has no capacity for human emotion other than experiencing perpetual, sheer terror. His ability to comprehend mathematical complexity, however, is unrivaled in all of human history. He develops new techniques to reduce decoherence problems of quantum computing, allowing the development of the primary AI in the story, Esai. His twin sister, Cari, exists for the sole purpose of providing emotional comfort and support to Sen.

Trying to develop such mathematical prowess in a single human proved to be extremely difficult, so they settled for twins: one with a mind that is barely stable enough to endure on a day-to-day basis and another with a deep emotional bond to keep him stable. The key to getting to this point was a legal innovation.

Creating human clones is mostly illegal. Experimenting on humans, clone or not, is incredibly illegal. So Promethia leverages paleogenetics to create Neanderthal brains. They aren't human legally, so there's no prohibition on experimentation, and they are similar enough to humans to help improve research.

Today, the world has caught up to my idea:

https://singularityhub.com/2018/05/23/living-neanderthal-mini-brains-may-reveal-what-makes-our-brains-special/

For those who remember, I posted this pic on the now deleted /pol/ thread. It was a brief timPost too long. Click here to view the full text.

9 posts omitted. Click reply to view.

62697a  No.204

>>132

>>126

Humans create or manifest in a synergy between mind and emotions, but the source is in the inner body. Its our link to "god."

You cannot get a machine to "create" like we create. At best, you can have an AI brain, with a human attached somehow, with them synergistic-ally working together.


bb945a  No.343

>>204

What if the intelligence is instantiated at the quantum level with true quantum computing? If that is possible, then it is technically not AI at that point, but man-made quantum intelligence.


07a709  No.416

Recent presentation by Svante Pääbo on paleogenetics:

https://www.youtube.com/watch?v=R1R8yrEGAgw

>Neanderthal sequences less frequent in gene promoter regions.

>About 30,000 SNPs and 87 proteins unique for modern humans, 3 genes related to cell division and neurodevelopment.


ea7eff  No.417

>>126

Liquid Math updates: We can now connect Neanderthal organoids to robots

https://archive.fo/l624F


a23243  No.418

Thank you for all your work!

I just fell upon your Twitter yesterday and have been diving into everything you've written since.




7937a4  No.375[Reply]

http://cultstate.com/2018/07/30/What-Is-A-Cult-State/

> Moore’s Law states the number of transistors in a dense integrated circuit doubles about every two years. By doubling the transistor density, the time to perform information labor halves. This results in the reduction of price for media creation, data analysis, and data collection while simultaneously increasing its distribution potential. This also drives an increase of the volume of domestic and foreign signal analysis competition against bloated, unaccountable centralized state-backed intelligence agencies. As Moore’s Law makes information processing cheaper, the blackmail network that keeps the nuclear paradigm stable can be discovered, evaluated, and captured in part or in whole by any upcoming competiton. Deterring this outcome has become increasingly difficult for state actors as processors, data centers, and data science techniques become cheaper and more abundant. The Hapsburg-dominated Church once experienced this same exact problem within the context of the printing press, and in a way, the transistor is the American printing press, but that’s an analysis for another day.

> The insurance file for Julian Assange is a massive collection of encrypted data. The unencrypted contents of that file have kept him alive through two administrations. This has made him a difficult person to make an example out of because I suspect within that insurance file is a significant number of video footage, images, and information about the blackmail network. Upon Assange’s death, his deadman switch will activate, and the key to unlock those files will be available to all who have downloaded the file. What options do central intelligence agencies have to thwart this trend?

5848fe  No.395

https://twitter.com/CultStateDotCom/status/1042173953798823943

> Canadian beatings will continue until the appropriate parties flip on each other.

> #thirdTempleDown

It's his 666th tweet

hrmmm




043ee6  No.384[Reply]

79ca40  No.387

>>384

There will come a time when my OTHER proximity to this situation will be revealed.

There's a whole backstory here that is fairly unbelievable. It's just not time to reveal it yet. :(


9cece6  No.392

>>387

Do tell.




93243e  No.156[Reply]

I do not consider myself a smart person. I have experience that has proven as much, at least academically to me. Despite this, I would like to think I have found a shred of wisdom for myself through my experiences, which has lead me here.

I know the follies of overstating my intelligence and the fallacies that can result from others doing the same. I think I'm smart enough to recognize the situation I am in and understand the absolute basics of infosec, metadata and the way megacorps use both. I am the most basic of analysts, someone who has a decent grasp of economics, a capable understanding of the human nature and a genuine interest in what makes the human mind tick, and more importantly, the way the human mind affects economics and technology by extension. Having followed the butterfly war thread(s), I found myself lost in some of the deeper technological meanings on some subjects, so I thought that a thread was due to consider how the situation could be more easily explained to those who have the capability, but not the prerequisite knowledge.

tl;dr what is the intelligent normalfag's role in the butterfly war and the skirmishes of gnostic warfare? This mostly applies to the "educated anon" role that will utilize and manipulate the tools that CS is developing to fuck about with the status quo once the phrenology manipulation tools become (relatively) mainstream.

2 posts omitted. Click reply to view.

7f4d1a  No.277

>>270

This is the new brown pill?


36be35  No.282

>>277

No, I'm just the most content person you'll ever meet. Continue to do whatever you're going to do, just know that it's probably being counteracted as soon as you do it. That's the nature of Fate.

Post last edited at

36be35  No.328

>>277

>>282

He's editing my posts because I am a bad faith actor who wants to spread demoralization instead of engaging in conversation.

Post last edited at

9c3983  No.385

>>168

I figured this would be the best place to ask something like this.

CultState, do you think there could be different types/branches of this Butterfly War idea? I know your main iteration is mostly focused on turning their own algorithms against them so that they're basically consumed by their own digital fire, but I was thinking about things along the lines of camouflaging oneself through either sarcasm or intentional misrepresentation so that while the algorithms don't pick up on it, people will and they'll be exposed to beneficial propaganda.

A lot of /pol/acks, either by will or by necessity, already hide their power level, and if you extend that idea even in normal, non-digital interactions, you can subvert the leftist concepts really easily. This is even more pronounced if you pose as one of their own, making them eat their own "deconstruction of ideas" poison to the point that it overwhelms their own politics.

I think similar ideas to the Butterfly War can be applied with considerable success in these typically normal moderated-by-consensus discussion spaces, especially ones considered "safe," as in almost exclusively leftist and often academic in the sense that when there's no other scapegoat for the left to push poison at, they'll consume their own all too eagerly


deb2cb  No.386

>>385

> do you think there could be different types/branches of this Butterfly War idea? I know your main iteration is mostly focused on turning their own algorithms against them so that they're basically consumed by their own digital fire, but I was thinking about things along the lines of camouflaging oneself through either sarcasm or intentional misrepresentation so that while the algorithms don't pick up on it, people will and they'll be exposed to beneficial propaganda.

Yes, of course. There will be millions of iterations and variations of this idea. The current investment structure of AI makes the entire surface area of entire nations incredibly easy to toy with. Sarcasm research has been done extensively. Double meaning is still a tough nut to crack, which is why my Incorrect Synonym attack works.

> A lot of /pol/acks, either by will or by necessity, already hide their power level, and if you extend that idea even in normal, non-digital interactions, you can subvert the leftist concepts really easily. This is even more pronounced if you pose as one of their own, making them eat their own "deconstruction of ideas" poison to the point that it overwhelms their own politics.

A fun side effect of the censorship wars is that /pol/lacks hide their power level online.

I'm slap in the middle of Hollywood in the belly of the beast and I -never- hide my power level in person. I've even been able to secure tacit support for eugenics in ground zero of this Leftist cathedral shithole. How? Try this out next time for fun:

> “Oh, you think seeking guidance for genetic health is taboo? Yeah? How many people with down syndrome have you had sex with? Exactly.”

I can't keep getting away with it, and yet, I do. How? Simple.

The more we deny our power level online, either by self-censorship, filter bubble distribution, or active suppressioPost too long. Click here to view the full text.




d2aa4a  No.226[Reply]

To help prepare for the shit that is about to hit the fan, I am implementing a dual canary scheme. This is NOT LIVE and is currently EXPERIMENTAL ONLY.

When this is made live, this thread will be deleted and a new one will be created with the final rules.

Experimental: Proof of Consistency

With the aggressive removal of the blackmail network that kept South Korea in line with globalist agendas, the sixty-five year war between the Koreas is starting to come to a close. This means the nullification of future loans (and interest) to pay for defense spending and post-war construction loans… and the internationalist banking elite are not happy. They will be preparing a worldwide multi-domain counterattack and Chan culture is in the cross hairs.

Chan culture is a vital part of the pro-national Trump infrastructure. Since Henry Kissinger, the Council on Foreign Relations, and Bilderberg have made media appearances in the wake of the Butterfly War problem, We’ve already seen an uptick in bot spamming and poor Markov mirroring on /pol/. These bots were trained on poster data in which, ironically, came from well-meaning /pol/ users themselves. As they labeled clearly disruptive posters as “shills”. This phrase is a special and organic classifier that flags the offending post as containing content as an accurate disruptor of Chan community. Those flagged posts have been used to train an anti-/pol/ neural network which is now being unleashed upon us and being fine-tuned for maximum efficiency. In addition to this, you’re going to see enhanced isolation and virality disruption campaigns. All of these tactics work in concert to ensure your next wave of meme war fails. If you haven’t been engaging in Butterfly War by now, this is a good time to start. It’s a surefire way to spread your memes under these incredible restrictions.

This also means targeting influencers. My honeypots are on fire these days and I’ve had to put Icarus on a kill switch. I suspect these attackers are not global observers because global observers knew I had a MAC whitelist on my routers. This means the partiesPost too long. Click here to view the full text.

9 posts omitted. Click reply to view.
Post last edited at

d2aa4a  No.370

>>287

come on bruh

there's no data there

https://i.imgur.com/QsYN6gt.png


d2aa4a  No.371

>>370

disregard that, i suck cocks

header was bad, fixed it


d2aa4a  No.372

>>287

https://i.imgur.com/gIqteK0.png

hrmmmm maybe the board mangled some characters?


d2aa4a  No.373

>>372

>>287

oh i see

you're a windows user, okay

lemme see what I can do


99616a  No.374

>>226

>muh proof of life

No one is going to murder you for writing javascript helper utilities and larping as a hacker.




307a60  No.247[Reply]

Battlespace Analysis

https://image.ibb.co/cOpPcy/01_of_12_Battlefield_Analysis.png

This is a high-level overview of the infrastructure, personnel, and operational behavior of the efforts being deployed to suppress, disrupt, distract, and infiltrate /pol/

The purpose of this analysis is to compromise the efficiency of neural networks and bots while forcing your adversaries to rely entirely on memetically-susceptible humans.

30 posts omitted. Click reply to view.
Post last edited at

ec6b25  No.330

>>317

Just to reiterate my previous point, not only am I a bad faith actor trying to demoralize the board, but I post three posts to make sure that all actual replies and comments aren't seen on the front page for new comers.

I pretty much deserve the way I'm being treated right now.

Post last edited at

280d8a  No.347

Until recently, nearly any input could fool an object recognition model. We were more surprised when object recognition worked than when it didn't. Today, object recognition algorithms have reached human performance as measured by some test set benchmarks, and we are surprised that they fail to perform as well on unnatural inputs. Adversarial examples are synthetic examples constructed by modifying real examples slightly in order to make a classifier believe they belong to the wrong class with high confidence. Rubbish class examples (such as fooling images) are pathological examples that the model assigns to some class with high confidence even though they should not belong to any class.

Myth: Adversarial examples do not matter because they do not occur in practice.

Fact: It's true that adversarial examples are very unlikely to occur naturally. However, adversarial examples matter because training a model to resist them can improve its accuracy on non-adversarial examples. Adversarial examples also can occur in practice if there really is an adversary - for example, a spammer trying to fool a spam detection system.

Myth: Deep learning is more vulnerable to adversarial examples than other kind of machine learning.

Fact: So far we have been able to generate adversarial examples for every model we have tested, including simple traditional machine learning models like nearest neighbor. Deep learning with adversarial training is the most resistant technique we have studied so far.

Myth: Adversarial examples are hard to find, occurring in small pockets.

Fact: Most arbitrary points in space are misclassified. For example, one network we tested classified roughly 70% of random noise samples as being horses with high confidence.

Myth: The best we can do is identify and refuse to process adversarial examples.

Fact: Refusing to process an adversarial example is better than misclassifying it, but not a satisfying solution. When there truly is an adversary, such as a spammer, the adversary would still gain an advantage by producing examples our system refused to classify. We know it isPost too long. Click here to view the full text.


e79913  No.348

CS - are you familiar with HLI / Three


307a60  No.362

>>347

This is a good find. I'm a huge proponent of adversarial training due to its implied biomimicry.

> Refusing to process an adversarial example is better than misclassifying it, but not a satisfying solution.

I wonder what happens when this is paired with knockout training… could be exciting!

>>348

I am not.


8d53a3  No.369

>>347

>Myth: No one knows whether the human brain makes similar mistakes.

>If our brains made the same kind of mistakes as machine learning models, then adversarial examples for machine learning models would be optical illusions for us,

Doesn't the second portion logically imply that optical illusions are the equivalent of adversarial examples for the human brain? It's important to note the difference between human object recognition and machine based object recognition but I think it's plainly obvious that human object recognition is also prone to certain types of error.

One personal anecdote that shows the weakness of human object recognition mechanisms (or at least mine) is my frequent inability to recognize a familiar object when looking for it when it is in an unfamiliar place. Example: I need a specific piece of mail so I start looking for it in my office mail pile. I'll spend 10 minutes looking for it all over the place only to give up, sit down at my desk, and realize it was sitting right there the whole time. My mind presumes a location for that object and combines the symbols of the envelope and the mail pile to forma the expected image of what I'm looking for (that envelope buried in a pile of mail). However, if that presumption is wrong and that expected image too strongly encoded in the mental process of searching for that object, then I am unable to recognize thaf object independently of the expected image. Basically, our brains (or at least mine) have developed highly efficient processes that allow us to identify objects based on logically deduced expectations of future events. This works great up until the foundational "training data" upon which those expectations are based is fundamentally wrong. I also think this plays heavily into the trait of "adaptability", the logical depth at which the foundational "training data" is based, the more flexible the process becomes.




d276a9  No.340[Reply]

Butterfly War Code Repository

https://gitgud.io/butterfly-war

This thread is for the discussion of the new Butterfly War code project. I have broken down all of the domains of work into individual projects, complete with milestones and high-level issues. Discussions about individual issues should take place at the issue. General discussions can take place here.

These repositories should be publicly visible , but you will have to apply to be a Member to each project in order to contribute.

These code bases will be open source.

62238b  No.349

Is it just me or is there no code in those repos.


d276a9  No.361

>>349

There is no code at this time. The project is being organized at a high-level to help anons (agnons?) get the 40,000 foot view.

Over time, code will trickle in. Right now, I am experimenting with candidates for data lake landing zone management.

https://github.com/Emblem21-OpenSource/wasabi-store-sdk

I'm currently writing this to get the process started. The –encrypt flag currently doesn't work, but that will get fixed so




84f0df  No.352[Reply]

Using Machine Learning to automatically detect antifa in Twitter

#antifinder:matrix.ordoevangelistarum.com

(Sentiment Analysis, Blocktogethers, CNNs and LSTMs, Emoji/Keyword Extraction)

Use Machine Learning and Image manipulation to flag (((Jews)))

#jewdar:matrix.ordoevangelistarum.com

(Facial Landmarks, image CNNs, Dataset Gathering)

Instead of using "subliminal" methods, one must first understand using ML in an easier manner to target the enemy's foot soldiers.



365001  No.319[Reply]



Delete Post [ ]
[]
[1] [2] [3]
| Catalog | Nerve Center | Cancer
[ / / / / / / / / / / / / / ] [ dir / agatha2 / dempart / doomer / vichan / vietnam / vril / wmafsex / wooo ]