Meme Warfare and Protection of the Republic
AI and the Meme Warfare that is upon us.
This thread is dedicated to fighting the Meme War.
Tips and Tricks and any bits of info that you have on how AI works, and how to thwart it to deliver our message to the people of the world.
The Cabal is losing, WILL lose.. this is our thread to talk about how we can DOMINATE the Meme War.
Okay so I started with the MetaData of the picture. It is the WHO WHAT WHEN WHERE and WHY of a picture.
It shows ownership of a picture (First Pic). (Second pic) I am using Adobe Bridge, you don't have to use Adobe Bridge for this, you can simply right click on a picture and from the drop down menu you can see "info" or you can go to "file" at the top of your browser window and should be able to get to the file info there also. Many ways to get to the picture/meme's metadata.
IF the MetaData was filled out for a picture to give "copyright status" to another entity it could help with thwarting AI, but I don't think it will 100% insulate the meme from being detected as /ours/.
I am hoping others join in and talk about the Hex editor and all. I will do research in this also.
I know that most don't meme, but those of us who do.. the creators need to protect our creations from getting censored on the webbings.
I DO believe the MetaData is important, even though AI will be "optically scanning" the pictures/memes.
Okay so I started with the MetaData of the picture. It is the WHO WHAT WHEN WHERE and WHY of a picture.
It shows ownership of a picture (First Pic). (Second pic) I am using Adobe Bridge, you don't have to use Adobe Bridge for this, you can simply right click on a picture and from the drop down menu you can see "info" or you can go to "file" at the top of your browser window and should be able to get to the file info there also. Many ways to get to the picture/meme's metadata.
IF the MetaData was filled out for a picture to give "copyright status" to another entity it could help with thwarting AI, but I don't think it will 100% insulate the meme from being detected as /ours/.
I am hoping others join in and talk about the Hex editor and all. I will do research in this also.
I know that most don't meme, but those of us who do.. the creators need to protect our creations from getting censored on the webbings.
I DO believe the MetaData is important, even though AI will be "optically scanning" the pictures/memes.
Filenames are important, show up in the coding of the page element that has the meme in it. This may seem really basic, but honestly we look at file names all the time "DOITQ!" comes to mind.. so this should not be overlooked.
This is going to be the first thing that AI will detect. It is important to have the filename correct, then the MetaData comes into play. AI has found the picture, now what is it about? It will check the MetaData to find out. Remember MetaData is the "Who, What, When, and WHERE" the pictures were taken.
The MetaData is where the picture was taken and any information about the picture that is available. New cameras will fill in a lot of data w/o your knowledge. GPS and the like. Most pictures have the camera, the ISO and all already filled out by the camera when the picture is taken. Serial numbers of the camera may also be added.. which could link back to you much like a MAC Address does on a computer. It is all data, the more data the easier it is for them to classify it.
This could be a good thing.. if AI is going after "human interaction"… or people tweaking a picture to make it into a meme.. we may need to take a picture of a meme after the meme is created.. then upload to websites. This would "erase" the human interaction portion of the meme/picture.
Someone suggested opening the picture in MS Paint and adding a few black dots.. if AI is going after "human interaction" or us tweaking the pictures, AI may flag the picture for that alone..
A thought is that meme generating services like https://imgflip.com/memegenerator may actually be working with the cabal and it is actually adding digital bits that the AI will look after to see if present, then it will censor because of those extra bits. Or any signature that a meme-generator site outputs.. I am sure that will be one of the first things it will look for, if the picture/meme was edited at a meme-generator site. This makes me think that if you can do the meme on your own computer, and control the MetaData we will be a few steps ahead of AI.
Digging on Rossetta AI right now.
from what the code is saying, it is looking for information about the picture/meme. I know that this is for another application, but you can apply what is stated here as a search for memes vs. the watching of people to suggest "things" that that person may want/need.
Like using a car for transportation only, OR using same vehicle to drag race on friday nights. Same platform (car) but used differently.
Okay, I read the articles, it can actually read text. So it is a text that has been placed on a picture.
I am not sure this Rossetta.ai is what he is using, but I can see what can be done with it.
So I got to thinking, it has to be apart of a picture. It cannot be added text.
In Photoshop you can create clipping masks. Take a picture, and with that it will clip wording out using a picture. Flatten the layers after that.. really doesn't take that long to do, a min tops, but you have to have Photoshop to do it.
Coupling that with the metadata.. I think we have a work around.
This is like the CAPTCHAS that are used.
I wonder if AI would catch this one though they probably don't allow US flags
I have decided to dig the shit out of this. Below is link to the whitepaper on Facebook's Rosetta Large Scale System for Text Detection and Recognition in Images
http://www.kdd.org/kdd2018/accepted-papers/view/rosetta-large-scale-system-for-text-detection-and-recognition-in-images
Maybe.. I d/l'd some DAZZLE CAMO from WWII they used on ships and it is hard to read.. which will be hard for Memes.. unless it is needed.. so they have seriously put a good block up to us, but we are Anons.. they cannot control the narrative, we will prevail!
The lettering is well defined in these memes. The program seeks out to locate text, probably has to see defined letters otherwise it will not catch it..
This is just like a captcha.. supposedly AI cannot crack CAPTCHA.. so to get around the software we need to figure a way to add wording like a captcha only legible to us.
One of the ways to make wording stand out is to put a black border around it. So when it is in a dark area, they white of the letters stand out. When in a light area, the borders allow for the word to be seen. So it is like the best of both worlds.. we all know good borders around words make for easier reading.
So no borders, and complicated textured text that does not allow for easy reading by computer. I am working on it.
I have some thoughts about memes in general. I'm not sure the full implications yet, but memes seem like a shorthand communication that can even be written by God or found in his word. Examples like "the handwriting is on the wall". The memes we make can communicate outside the internet just like seeing a Q on a bumper sticker means something to us all now. Early Q drops had embedded messages like "C_A is here". We need to use all aspects and dimensions of meaning to evade the censors and we can develop comms that work IRL. Would an AI know what 86 means?
Very smart! Someone last night was talking about L33Tspeak but you have to know that to get it.. still has to be simple enough people will understand it
the words have to be easy to detect
We should know our enemy..
I present to you the first of 3 of the bastards who are in charge of Rosetta:
Fedor Borisyuk
This dude is bad news. Ties to actual Russia also.. will post about that in a minute.
part 2.. this guy is a fucking GEEK! Evil Geek..
He was apart of RuSSIR 2009 is "The goals of the Russian Summer School on Information Retrieval (RuSSIR) are to introduce students to the spectrum of modern problems and methods of information retrieval, to attract the attention of students and young scientists to the tasks of information retrieval, to create a platform for informal communication between representatives of science, higher education and industry. "
The below website is in Russian. I translated the page (posted pics)
http://romip.ru/russir2009/
On to douchebag #2
Albert Gordo
"think [nuclear] football"
Albert Gordo.. just looking at his links on his linked in.. holy fuck..
ICFHR 2014 COMPETITION ON HANDWRITTEN KEYWORD SPOTTING (H-KWS 2014)
http://www.icfhr2014.org/wp-content/uploads/2015/02/ICFHR2014-H-KWS.pdf
10th IAPR International Workshop on Document Analysis Systems
March 27th - 29th, 2012
Gold Coast, Queensland, Australia
http://www.ict.griffith.edu.au/das2012/awards.html
The PASCAL Visual Object Classes Challenge Workshop 2012
12th October 2012, ECCV 2012, Florence, Italy
http://host.robots.ox.ac.uk/pascal/VOC/voc2012/workshop/index.html
CVPR 2015 Awards Page
http://www.pamitc.org/cvpr15/awards.php
3 CVC PhD thesis have received the Outstanding PhD Award from the CS department of the UAB
http://www.cvc.uab.es/?p=1927
Viswanath Sivakumar
Rounding off with #3.
No picture on linkedin.. going to try to find one somewhere else.
nice that I have been working on this, and all the sudden I somehow get a ban.. from July 31, 2018? WTFF?? Yeah.. I don't know.
I have a couple more things I am trying out on memes.
I wish Q would say for certain if they have the ability to actually detect Memes. Why I say this? Their white-paper says they DO have the ability to read wording on a picture, from a hand written message/word to graffiti.. words and letters put "on top" of a picture like a meme.. wording over a picture. I will just search for what they said and post it above.
Here is a Susan Rice I was messing with a minute ago. Trying to hone the craft it is a fine line between not being able to read it and totally being able to read it.
As far as the ban goes.. i have no fucking clue what that was all about.. but whatever.. catch most flack when over target.
Q if you happen to read this, shed some light on what they have available or what they can legally do with it, if that has anything to do with their not being able to detect memes.
The Russian guy they have working for them on this.. that dude is INSANE.. he is into some serious AI shit and has a mile long resume.
Maybe you have more than you know. The paper ref’ed the open source Detectron. Pics also get normalized down to 800x600pix, so maybe the knowledge of data loss from really large images could mess with this process?
https ://github.com/facebookresearch/Detectron
Brainstorming a couple of things:
Leet speak.
Clipped words
pattern fills in text masks
captcha fonts
Old horrible fonts
We will have to rely on the human mind to fill in what computers can't
backwards writing
loading false exif data
I cnduo't bvleiee taht I culod aulaclty uesdtannrd waht I was rdnaieg. Unisg the icndeblire pweor of the hmuan mnid, aocdcrnig to rseecrah at Cmabrigde Uinervtisy, it dseno't mttaer in waht oderr the lterets in a wrod are, the olny irpoamtnt tihng is taht the frsit and lsat ltteer be in the rhgit pclae. The rset can be a taotl mses and you can sitll raed it whoutit a pboerlm. Tihs is bucseae the huamn mnid deos not raed ervey ltteer by istlef, but the wrod as a wlohe. Aaznmig, huh? Yaeh and I awlyas tghhuot slelinpg was ipmorantt! See if yuor fdreins can raed tihs too.
detectors are going to need a sharp edge to find the text lines to know there is text.
feathered masks would help hide. B/w images, blurred words
how many colors can computers see? How many beyond 256? Humans can see 6 million.
Best I can do with the crude tools available to me on this timeline.
The Big O on the down low… #44GITMO 🌴
Did anyone happen to notice this morning, certain images (profile pics, shared images) on FB were loading as an X? You click on the x to lead to a blurb about it containing text and potential fake news/misleading information. I tried to get a screenshot but was too late. Must have been a text or blip.. preparing for thursday
Yes I can read it :)
Damn no shizz huh??
I have been on the regular bread today, but I have thought about
BURN technique in PS
STAMP technique in PS.. where you can mess with the words and then just stamp it and flatten when done, then save..
Important for the MetaData.. so huge.. we need to find some addys for DNC and such.. I think the one I provided above is that.
I will work on this more tonight, should have today.. I was kinda feeling like I was talking to myself.. glad to see people interested in it, and damn! Go and do that so we can see the X! We are on the front freaking lines Anons!
Yep sharp images, I was thinking about doing an edge around all the wording, go into "CUT IMAGE" or whatever it is called (PS is all broken up into diff programs now.. no likey) and then you can go around it erasing the edge, but you can control the "softness" of the edge, crank that bitch to 100..
Was also thinking a picture of sand or rocks.. whatever it is.. slightly individually making it diff coloring or clip a word into the top layer, slightly color it and overlay.. what i was doing above already.. but different.
I know that there are different tools you can use, I am just getting back into PS. I had CS6 Master Suite and had a MBP get warrantied and lost the key for the new laptop and refused to "rent" the software from PS out of spite and CFT (Cause fuck that)..
So adjusting to new stuffs.
YES YES YES!!!
What about leaving letters out of words? Humans are really good at filling in the gaps.
Sorry, not tech person, but do meme. Can I sit at your lunch table?
you just read my mind.
All I know is that a pic is just a pixel map. each pixel has a color code. alter pixel color code of pixel map to make letters. Probably script it using something like this:
https://www.linux.com/news/advanced-image-editing-command-line-imagemagick
Unfortunately, camouflaging text drastically reduces impact of memes.
I can not afford PhotoShop… so I use Gimp https://www.gimp.org
What about just inserting junk words every bother word in a different color? or redundent words? or again every other word backwards… is there a capture based program that will allow reg people… ie me to make vaptures?
This particular filter looks very promising because it breaks up edges of text while leaving the text readable.
This is gimp 2.8 and I forget if it's a standard filter or something I installed a few years ago. It's in the filter category subcategory Artistic.
I believe the OCR is looking for edges and this defeats the search for distinct edges. But if they use a blur and rescan they might be able to read the text.
P.S. Before this bread was made we posted lots of examples of things we were trying in Memes30 over the past 48 hours. Have a look over there are maybe you will get some ideas.
Also try curved text (haven't tried that) much yet.
Also make a selection from text, then use Select / Distort to change the outline.
Camo text was made that way.
Make a text layer, then use blend filters to layer it over existing colored background so that the text undergoes a color shift.
Had this idea and posted it in the general. Instead of focusing on bypassing the AI, try to destroy it, or at least render it inoperable. Give it logical fallacies to solve, math problems with no solutions, get it to try to solve Pi or send it into an infinite loop.
Godspeed Memefags,
WWG1WGA
That's not going to work. If it's programmed to examine images for probable text, nothing you post is going to be interpreted as a logical fallacy or a math problem. It either looks like text or it doesn't.
I used gimp way back in the day, I am all for any software, we just have to work it out. I will be on later, I have lil one.. we are going to go do some fun stuff.
What is crazy is I got banned yesterday, my IP did.. then I thought that was weird, said must be over the flack.. what happens is some chick from NBC sees the thread and writes an article about it.
If any of you think this is a game, that we are not fighting a war.. they tried to ban the creator of the thread.. they wanted this stopped.
I have an idea that will turn a meme into something that cannot be detected. They wont be able to.
Their AI meme generator is a joke. And their ability to see text is limited to defined text. If you flatten the image, then even take a picture of it or.. that is where I will have to figure out and discuss here about erasing the fact that it was edited away. Like a stock photo.. they just can't do it.
And here is the deal, yes.. it makes the meme harder to read, but we have to adapt if we are going to do this. I think it is a sign of the times, at one point shoes went from moccasins to laced up shoes.. things change, we have to be fluid and adapt.
3 more text obscurations using Gimp 2.8
Reduced contrast with background colors.
Broke up text outlines.
Still human readable but not so much.
This style might be good for confusing AI? An anon in the research board posted it.
>>3078420 Yes that is hard for people to read. High probability it could deceive a program.
Anon posted this in Memes30.
Looks like this font will work!!!!
Q underscores NSA.. so that means that only NSA can figure out memes.
From the whitepaper they talk a good game. I printed it out and read it, have it here.
So Facebook must be trying to position themselves to fake us out.
I would not stop doing this, I am going to continue myself, tonight when I get back. But I am pretty sure Q is saying they do not have the capability to do that.. to detect memes correctly.
The article states that they can pull it out (the stops) when needed most.
Hey Facebook.. how you gunna get your AI to read Captcha? If this is Tit 4 Tat its on motherfuckers..
Spend your $$$ cause we are doing this on our dime.. and we don't need PHD's in AI to come up with the ying for your yang.
And since you are watching.. SUCK MY COCK I have ALWAYS hated your product(s)
YES! and title it and metadata it in the way that looks friendly..
Here is the deal, will have to have multiple multiple multiple memes so that when they flag a certain "owner" you have more out there.
If this is a game, gotta play it. Q says we have forgot how to play..
Not true.. we have just begun
we could use twitter multi-pic grouping to overlap text over a two or four group image post to create a full image out of four equal parts that couldn't be A.I. scanned individually with any success.
Instagram posters did this for a while to create collages and murals with their albums. At least would buy us some time and force them to change their system to beat it.
A.I.?
MORE LIKE..
GAY-I.
GIMP is the shit
copy an image from a webpage
go to gimp - file -create -> from clipboard
put some text on the page
change the layer Mode to "Difference"
filters -distort -> curve bend
profit
V
E
R
T
I
C
A
L
T
E
X
T
Should be unrecognizable.
easy for optical character recognition to strip out the whitespace and it is only a word
>http://www.gbif.se/digitisation-symposium-2014/13_Brun_OCR.pdf
Ok. How about tiny mp4s with static images with sound bites that say the text?
Inkscape is an awesome vector editor like free illustrator. Been using it for a couple years and like it better than illustrator now. You can export png from it. Text on a path, filters, ready for memeing it up. Also check out this site https://lingojam.com/TwitterFonts
Artificial Eye, Artificial Vision: How does my robot see? - Professor William Ayliffe
https://youtu.be/YltPTIFq_Hc?t=2256
I fucking love this
>A thought is that meme generating services like https://imgflip.com/memegenerator may actually be working with the cabal and it is actually adding digital bits that the AI will look after to see if present, then it will censor because of those extra bits.
That's a really good point.
What's the easiest way to strip/alter the metadata?
I had an idea that I wanted to share. This is just a rough draft, but I wanted to take a movie quote and alter it to make something a bit spooky. Thoughts?
Right click an image, go to info.. fill in the fields 👍
These are really good. Well done anon. This challenge could really inspire some nice aesthetics.
Very cool, think we should write like that on the general board to screw with their linguistic analysis programs. Is there a program to generate text like that?
Problem with using that text in memes is that memes need to make an immediate impact, and this text introduces a challenge to the audience that they most likely will not understand the purpose for. Thus the potential impact of the meme is reduced.
Some block letters someone posted on Sunday.
Interesting about the new article that Q posted today is that they mentioned that each "fix" could take the machine learning a few hours to a few weeks to accomplish.
If we're all doing our own manipulations separate from everyone elses, they're going to have to run a "debugger algorithm" (not sure of terminology) on each manipulation.
Few minutes to create vs. few weeks-few months to resolve sounds like pretty good odds to me.
Easy peasy, thanks. 👍
Clise the media. Abolish masonica. Never with the masonic oaths broken current year.
block letters…
like newspaper/magazine clippings for hostage notes
we could take pictures of the text made from letters cut out of magazines
I like this idea. Their biggest obstacle is bandwidth and server constraints… so obvious answer increase file size drastically with MP4, or something uncompressed. Live stream memes?
Massive organized dumping of HD 4K Video Memes will turn those servers into a second sun.
what about GIFs? as far as I can see the system downsizes to 800 maintaining the aspect ratio. is it taking a screenshot or the entire media into consideration? if only we had this software to test hands on
>Meme Warfare
I would think that screen shot would overcome this. Also think that it checks the MD-5 hash to see if it is has seen it before.
I don’t think gifs would work because they are just sets of images that would have the text embedded on at least some of the images and therefore the algorithms would pick it up on those images and then cordon off the whole file based on that images.
Another idea — I wonder if there is/are any special characters we could add inside the line of text (or in key words within the text) that would mess with the algorithm? Just a character that would trip it up so it couldn’t read the word?
some experiments
moar better…
Here is a work in progress
Nice picture Q
I didn't realize that Peacey was a Jew…
https://twitter.com/LoveLig66141315/status/1039227865185677313
The more variation we try, the more they'll have to retrain the recognition algos.
Anons who are trying >>3079160
colored text over a different colored background, I don't think that will work well. All they have to do is pick out the green text from a non-green backround, easy peasy.
I believe the small text on this one >>3079627 will be difficult for OCR to read. But the hashtag? Can we do memes without hashtags? They can instantly read and ban based on readable hashtag.
I see a lot of progress as anons are trying different things. Good creativity and energy, memeanons!
complicated
KISS
by the time you break one system the bad guys create new stuff. Simple pass EO to block MSM from censorship of any kind. While MSM in court fighting the EO we string into action.
I was wondering if the following techniques would work, I don't have time or skill and I'm not American so I'm not sure if Memeing is interference? Anyway…
1) Make large words made up of smaller words. Use Dem0crat buzz words in the small words so that they pass the filters.
2) Cut the LETTER of your WORDS into random blocks using lines that change in size. Think of a slice through a letter that is bigger on one side than other. Off setting chunks will mess with pattern recognition.
3) Use dots to write your Words intermixed with lines and shapes. You're trying to deceive something following lines and building shapes from pre-scanned patterns.
4) Have headlines like VOTE BLUE WAVE on the top, followed by these word techniques below with your POWERFUL message. The idea is to have them PASS the filters with their probably CHOSEN key words.
These ideas are based off how an AI would be developed to interept Memes. They look for structure.
If you can mess with how the SHAPE of how your word are formed, you FORCE them to react to your variation. Reacting to your variation takes time and money and many of your messages will get through.
5) Use Wrods taht keep the lsat lteters the smae. (Least efficient)
Thanks for chatting.
Quads confirm. Nice memes!
>Can we do memes without hashtags? They can instantly read and ban based on readable hashtag
right! the idea here is to use popular left hashtags (#resist, #imwithher, #abuseofpower, #metoo) that are easy for the OCR to find so it will classify the images as pro-left memes so they go through even with the Q message
think like some shitty software written by facebook import programmers
face = polosi
tag= #resist
face = clinton
tag= #imwithher
stereograms or 'magic eye'
code is here
https://github.com/peeinears/MagicEye.js
This meme and poem is driving me crazy!!!
Take it away!!!
i fix one thing fuck up two…
i think i have it right this time…
apologies… kek