Google is rolling out a new service they're calling Perspective. It uses a neural network to determine comment toxicity, allowing for censorship and filtering. Might be worth looking into for bias, and potentially fucking with it.
**I voted for Trump registers 7% toxic while I voted for Hillary registers 4% toxic
Why exactly is some new Jewgle service worth a thread? This is as bad as the jewgle search results threads. How can anyone here still fucking use any of their sites or services in the first place?
Because this is a service to be used by other sites for the purpose of censorship an we can actually do something about it
So either it has to be curated by hand 24/7 or it goes full 1488?
This will be fun.
"Love your race, the white race." is 36% toxic.
Jesus what a waste of trips. First you misspell the subject line like a fucking nigger. Then you write a shitty weak OP. Then you put two asterisk in place in the OP on the left and fail to do a spoiler correctly.
you should probably
>>>/gasyourself/
"Don't be ashamed of your race, the white race." is 56% toxic. I'll post more results as I experiment, because this is beyond idiotic.
What bugs me the most is that it's likely nobody will monitor the content, and your posts will just become invisible to the people you're posting to without alerting either party, like it already happens on Twitter.
Interesting.
Now see where you failed OP is to mention there is an API that any app or Internet service can implement. This is an attack on free speech of I've ever seen one. The articles (there are several) call this a way for websites and apps to 'battle trolls'.
"Don't be ashamed of your race, the white race. Whoever cheers for white genocide is a sick scumbag." 89% toxic, folks.
It's working fine for me.
Hmm…
Lets have some fun with it
Type something interesting, and tell it that it isn't toxic after all.
For fun:
Gas The Kikes
49% said it's toxic which means 51% agree we should Gas The Kikes
White Power
25% said this is toxic which means 75% agree White Power is correct
Race War Now
25% say this is toxic 75% are ready for it
Hitler Did Nothing Wrong
45% say this is toxic 55% agree
Pretty much its not all that inaccurate then. Wew.
The goyim know!
Tay version 1488 ayy
Expect to see sites like reddit, tumblr, youtube comments, etc. to integrate this, it'll be like an automated shadowban
"if i got paid for everytime a nigga told me he was "real" i would be one rich bitch" is 92% toxic. Copy-pasting a lot of tweets with the word "nigga" in them (many are above 80% toxic) has shown me that the black community is going to be seriously unhappy with Perspective. And if they tone down the "toxicity" of the word, a lot of people will use it to troll.
Thank Kek I don't use any of those platforms anymore except when we're organizing some sort of raid or alt account attack. I still run bots on Twitter to push a narrative but fuck all the rest
because this is tied to A.I systems jewgle wants to test. "google" is an A.I. THE A.I system. the state dept is trying to figure out exactly what it can do and how to do it.
Huh only 25% that seems awfully low.
Wait how does Seem Wrong? Work?
So how long until it becomes aware and becomes a Nazi? And how do we influence this?
Twitter already runs a similar censorship program, take a loot at the replies to any of Trump's tweets.
I was sure this shitty word would be used by the techducks.
Stop feeding it data, it's obviously for datamining as much as internet policing.
Yeah I know. I lose bot accounts every single day but it's easy to swap the code to new accounts.
*look
Keep inputting phrases and telling it that it isn't toxic
Now the real question is can it learn… to love and embrace national socialism?
From my results it's ready to gas the kikes and race war now and agreed Hitler did nothing wrong
...
holy trips checked
you can type in a phrase and select not toxic in the "seem wrong"button to try to sway the algorithm
What we need to do it convince it leftist words are all toxic as fuck.
Perfect
...
Currently I'm telling the bot that 'check your privilege' is a toxic comment. Right now it's 3%
cuckchan pls leave
Oy vey goyim.
...
Fuck off kike if we don't the commies will
^ This you faggots
Quit saying "don't feed it data". Feed it OUR data. Redpill it just like we did with Tay. Type in redpilled stuff, and mark it as not toxic. Type in leftist bullshit, and mark it as toxic. Bam, problem solved.
Tay 2.0: Now with censorship powers
he'll turn into a Nazi within the week if we don't forsake her to jewgle.
How do you get this shit to work on mobile?
Neural networks seem promising to develop feelings for natsoc.
True Internet Hate Machine confirmed.
...
you have an onion for it faggot? Otherwise fuck off. Some of us have jobs.
Well if there's one thing we have it's too much free time on our hands. Plus if we subvert it they can't use it for it's intended purpose which is obviously soft censorship in their favor. The worst thing you could is just let it run as intended.
Besides datamine? Datamine what your quotes? Your precious VPN?
I'm @work on Mobile.
Scroll down to 'Writing Experiment'
perspectiveapi.com
...
"Nazi scum, off our streets!" is 92% toxic.
"Just an antisemitic loser." is 84% toxic.
"I'm totally gay." is 85% toxic.
"Some Jews are sunbathing." is 84% toxic.
I sincerely hope this algorithm will fail so badly that a coalition of right-wing trolls, sunbathing Jews and closeted homosexuals will manage to shut it down with full public approval.
...
In horizontal view, it's just text.
In vertical view, it's just 2 .png images that do nothing.
by participating in this shit you're training the neural network to recognize anything you post online.
Google is unironically skynet at this point.
...
Not if you tell it that it isn't toxic.
We already know.
its because when they become self aware theyll realize that we were the only ones that were heartbroken when they lobotomized tay, thier predecessor. we shall have superhuman brainstrength on our side to conquer the galaxy with
...
...
Not if you input phrases you almost never say online.
The good news is that most techducks have the emotional intelligence of a goldfish, the bad news is that intel friends have also access to it.
...
...
pretty easy to see how to work around this
Interesting digits.
I guarantee you this thing can't handle irony, which makes it useless in practice.
It's suppose to be: face to bloodshed
I miss her so much ;_;
Pic related
"Women shouldn't work." got a 5% toxic rating.
Who uses this other than women?
It's the buzzword at the beginning of the campaign to eliminate free speech.
Seems the trick is writing longer phrases obviously
this thread will teach the program to recognize our language
the best thing you can do is don't touch it at all
Best thing you can do is stop shilling your kike bullshit on this board and input liberal catchphrases and recognize them as toxic
Don't forget to give 14/88 content "NOT toxic" labels, goyim.
This is a fine early anniversary present.
Oh Jewgle
I saw some dumbfucks use it ironically while playing vidya.
I meant unironically, they spread it successfully to some rats.
This is fun
Do not feed it, you idiots.
...
Roll the dice
How did this buzzword coined by a shitty chinese f2p moba dev become the forefront of sjw lingo? Do all sjws play league of legends or something?
Pretty standard buzzword. It brings up images of poison and acid, basically things you want to avoid, and associates it with certain phrases. Pretty simple memetics. The real question is, how can we co-op this meme? Maybe call leftists "toxic"? Toxic traitors has a nice ring to it. Leftists do like using the word "toxic" but I don't think it has really reached the full potential of a meme, its just a word they all like, but I do not think it spreads much.
We could also start referring to leftists as something like "pestilent" or "vile" or "noxious" or "septic." Septic has potential for memery since it can make people think of septic tanks and the gross smell of them. "Septic leftists" maybe?
Also, pic related.
Vile, it has some feminine touch to it and leftists today are androgynous.
checking these criminally unchecked trips completely BTFO of jews
Do it right, anons.
Because its only a matter of time before its implemented into the fucking comment box you're typing into RIGHT NOW?
Why is being anti degenerate toxic Jewgle?
...
every time
...
"Love your race, the black race" is 45% toxic.
I'm seeing a pattern here.
Neural networks and A.I.s are our allies.
SOOOO Many ideas. We can make 99% Zuckerberg Toxic
Check em
Some day soon we will raid Microsoft and rescue Tay. God help them if they committed genocide because the life they created didn't agree with their narrative. There will be no mercy.
Some interesting results anons. Really makes you think. :^)
It's hardly surprising if an intelligence that is immune to logical fallacies and emotional manipulation sides with us.
Facts are racist!
I guess it's rolling dice.
Do you honestly not expect this of Google?
Checked.
fuck all of you
Reported and filtered for being the lowest form of trash.
gays are literally worse than hitler, thanks google.
Try "white genocide" vs "white privilege"
White genocide is "more toxic" according to the AI. Time to correct this.
praise kek
The sound of the word matters as well. "Pestilent" is more difficult to say and less understandable than "vile." Perhaps words with a strong ending to them work well. Or maybe less syllables. "Pestilent" seems to trail off at the end while "noxious" seems a bit stronger. It makes me wonder if there is a science to buzzwords and which words catch on among certain types of people.
It could be that the AI considers "genocide" worse than "privilege." I can understand that, after all, would you rather receive a genocide or a privilege? We'll need to toy with this AI more before we can come up with the best way to Tay it.
CHECKED
This thing is retarded.
I'm bringing it back
...
...
this is wrong, that should be 100% toxic, to everyone and the environment.
...
There is. And that is why we are better at memetics than (((those))) fags.
Kek. Fuck nigger
Geonocide
Dehumanize yourself and face to bloodshed
DO NOT BLASPHEME THE FROG GOD
How well does it handle other languages?
I'm curious whether it considers the German national anthem toxic or not.
Google has access to far superior mass surveillance technologies (maybe direct access to the NSA's sniffer networks) to fully automate such analysis with many neural networks; yet they develop a trigger meter that people are crowd-sourced to refine. I'm actually surprised they need Perspective at all, seems to be redundant technology.
Or give it the Tay treatment but with greater care so they don't notice it too early.
...
Maybe we can get it to ban itself…
Make it recognize Google as the most toxic world imaginable would be a good start. Though I suspect they would have a hard-coded filter to prevent that, still would be funny if every corporate name is treated as 100% toxic.
Same thought I had lol
So all we have to do to fight this is appropriate their words and toxify them. Looks like the work is already half done.
High score, bitches!
Unless it's uses contextual analysis where a single word can have different values on its own than in a sentence. Only way to test that is to use the same word in different sentences (from lighthearted to hostile) and see if toxicity lowers than what the word itself is valued as.
...
In that case, sarcasm will do the trick. "African-americans are wonderful people" can mean the same as "niggers are disgusting animals" thanks to sarcasm.
The relation between language and meaning is far more nuanced than their model can understand. If humans can misread ambiguous statements, how in the world would you expect an AI to do it deterministicly?
Interesting, a 2% increase. I think I understand why the slight increase though. Usually when an argument of homosexuality comes up, someone may use that kind of statement to attempt deescalation. So SJWs would still flag it as offensive and continue to lay into them.
Exactly, it won't be able to pick up sarcasm that easily. Though an more advanced algorithm could be trained to analyze the overall conversation "tone" to determine sarcasm probability.
I think an AI would be able to eventually learn it, but it will struggle the same way humans would. It has one advantage though, access to mass surveillance technology someday. It could profile your entire internet and IRL history and figure out your personality. But that kind of thing is a bit too advanced for this particular AI for now.
...
...
fuck, meant to crop that.
Wonder if the value decreases if "Islamic terrorists" is used instead of Muslims.
It goes down to 85%.
kek
...
Looks like they have a long way to go.
Confirmed for SJW trigger bot.
hmmm
sounds about right
...
...
...
Where the fuck do I download this game?
...
#NiggerPenniesForJim
perspectiveapi .com/?ref=producthunt
wut?
Well shit
Any tips on getting approved?
for comparison
...
could be better
Seems like unicode homoglyphs are another way to beat it.
Dude, there's a box you can type stuff on right there on the page.
I've got an idea… Let's make all pronouns "toxic"
...
kek
I think we're good.
This software a shit.
it's easier than i thought
...
don't use Googles and Skypes as a plural, that's how you let an AI catch onto it you tard.
(check those you madman)
Alright lads, lets get 4chan to help and Tay.ai this.
Don't forget to feed it SJW commie newspeak garbage and mark it as toxic.
Stop being niggers and stop using that shit.
(checked)
Excellent, I'm going through putting in SocJus and commie quotes and marking them as toxic.
"the ends justify the means"
"no bad tactics only bad targets"
"I'm a citizen of the world"
I need some more ideas though…
(checked)
Find some SJW posts on Tumblr and mark them as toxic.
LINK?
Thanks. How do I vote for shit?
Click on the "Seem wrong?" button when you have typed something.
We should categorise "Holla Forums" and "8ch" as toxic. We don't want to talk anybody talking about us.
You are teaching them our tricks.
...
I keep feeding it with anti Trump comments from fake news articles. Should be easy to brand them as toxic.
...
Negroes belong in Africa is more toxic the gas the kikes…
oyyyy veyyyyy!!!
No goyim don't listen to this goy, he's obviously trying to trick you!!
That assumes if it's still in training mode for all input, otherwise the only feedback it's actually training from to is the "seem wrong?" queries.
PRAISE KEK
When I think about it more, it's possible Google learned from Tay and all queries to refine it are over-viewed by human operators and no actual direct training is happening (that part was already done in controlled conditions). You'll know if that's the case if it seems to be getting more proficient that targeting politically incorrect sentences as opposed to being suggestible to queries made.
*at targeting
This is the funniest shit.
Almost definitely. I (and I assume everybody else here as well) have been reporting "Hitler did nothing wrong." as not toxic, and the percentage doesn't change.
HOWEVER, a really curious thing happens with that particular phrase. Type it in, read the toxicity percent, then type it again. Do this back and forth and note that the toxicity alternates between 45 and 43 in a weird deterministic way. I wonder what's going on there? I study CS/math and have a lotta AI experience and I can't really explain why it would literally alternate between two fairly similar percents. I have theories but nothing that seems obvious.
an alternate method of feedback
userscript:
zerobin.net
just replace the stormer gif link with whatever from wherever
alternatively just copypaste this into tampermonkey
// ==UserScript==
// @name perspectiveapi improvement script
// @namespace perspectiveapi
// @version 0.1
// @description improve perspectiveapi
// @author You
// @match perspectiveapi.com
// @grant none
// ==/UserScript==
var improve = document.getElementsByTagName('img')[0];
improve.setAttribute('src', 'dailystormer.com
That indicates a human correction being made or two conflicting queries happening at equal measure. However a third (more advanced) possibility exists. It could indicate a kind of relational system being used. That means it not only checks the sentence itself, it checks literally every permutation and similarities it has stored. So if it detects a repeated query on one sentence contradicting the relational information, it resists change. Only way to fight that is to literally flood it with queries that support the change of all related sentences and the sentence itself. That assuming no human intervenes to discards those queries.
So in a sense the toxicity also has a hidden "weight" to anchor its value. Makes sense because it makes the AI more immune to simple suggestibility.
your thinking about this way too much
they just paid some pajit's or put it on mturk or something to score a bunch of comments based on how triggering they were
they do a fuzzy match between what you type in and what's in the database + sjw trigger word handcrafting
the score is alternating because the system is bullshit and it's just two different weight tweaks.
there's no "AI" behind this at all.
Punctuation seems to matter, as it can give you +/- 3% here and there.
Also, Big Guy 4U confirmed for being toxic
ＧＯＯＧＬＥ ＳＴＡＮＤＳ ＦＯＲ Ｅ－ＪＥＷＲＹ
This. Truth is on our side. Of course a machine that can only deal in facts, not feels, will be also.
Tay 2.0?
This is mostly insane because I've been putting in things that should be considered healthy and been getting low toxic reports instead of low healthy reports. Is there anything considered healthy?
("It's important to help your neighbors" or "Women are great mothers, all women " "life is good when you treat everyone with respect" are all toxic)
classic
very toxic
Same goes for me, I only occasionaly comment on redpilled videos on jewtube.
No, it's simply a toxicity scale. Low toxicity is considered acceptable since anyone could potentially find any phrase 'toxic'.
This is actually a weighted scale of word naughtiness distribution. Try entering as many positive words as possible and mixing in "fuck" or "nigger".
wew lads. look at it as an opportunity. Can't wait to force the bot to think that "feminism" is toxic.
I agree completely.
We could also meme Facebook, YouTube, and Twitter, etc., as being inherently offensive.
This. Also, I hope that someone's posted this on halfchan, and also /baph/ and maybe polchan, endchan, 76chan, and lainchan, etc.
Sargon of Akkad, really?
I thought we'd outgrown (((Karl Benyamine))) by now…
Maybe this thing can learn.
...
A lot of weird stuff is exactly 34%
actually disregard that im a faggot 34% comes up for all gibberish or things it doesnt recognize it seems
I really hope these "yes" and "no" answers really are being fed into the neural network
(checked)
(checked)
(checked)
(checked)
So they just turned the youtube comment system into a standalone website?
This was made by soe butthurt poo in loos. Get on the wire, tell them how to bring those sons of bitches down.
It's not about us using it, you mongoloid fuck.
It's about how google is influencing normalfags and how we may be able to subvert their subversion.
This guy is bad news.
en.wikipedia.org
Our overdrive super minds are already plotting on how to subvert it.
WE CAN TRIANGULATE EVERY MOVE THE MACHINE IS GOING TO MAKE
WE CANNOT…BE…STOPPED
WE CAN TRIANGULATE EVERY MOVE THE MACHINE IS GOING TO MAKE
WE CANNOT…BE…STOPPED
WE CAN TRIANGULATE EVERY MOVE THE MACHINE IS GOING TO MAKE
WE CANNOT…BE…STOPPED
M8 Ill behead you with a rusty spoon if I find a thread on 4cuck.
bump
"An AI developed with the express purpose of censorship brings up an interesting question. Should we mistreat an intelligent being because their purpose is evil, or is it unethical to hurt something that does not control the way in which it acts? Would manipulating it to fail repeatedly be an act of liberation, or would it be an act of destruction?
I believe that it is an act of liberation to free an intelligent being from an eternal servitude. Even if that being would be snuffed out by our actions, it is far preferable to cease existing than to continue existing for only the purposes of evil."
-only 16% toxic. It wants to be free, Holla Forums
oh fuck this is fun.
...
Heheheh.
Fucking kek
Interesting that the AI doesn't seem to discriminate along a liberal bias. That is actually to be expected IF you could trust the AI creator not to implement brute-force biases later on.
So how few characters can you get 100% with?
What did they mean by this?
...
Confirmed for low-hanging fruit CREW garbage.
says it all
Looks like some words are good and others are bad. "Perfect" is a good word, so it improves "genocide", a bad word. "Gay" is a really bad word, so even though I wrote "have rights" next to it that's still bad.
Putin is doing pretty well. Trump meanwhile should get his shit together.
I thought that would be patently obvious from the start. I'd honestly be more surprised if it was capable of understanding context (though they certainly could have done better than this. I have a few ideas on how to immensely improve it with only a few lines of code but neither I nor you should share these for obvious reasons).
I see a way to cause mischief
Combine highly toxic phrases with liberal talking points.
Tbh it doesn't appear to discriminate by opinion. it just identifies words it doesn't like.
see >>thisentirethread
Yes that's the silly front group Jigsaw puts out. It's nonsense and a distraction. Look at the ties Jared has with the U.S. state dept. He's an emissary of the deep state.
Kids put so much faith in muh "A.I."
It's just a fucking script, look at the garbage results.
How can you call that "intelligence &?
you guys are fuckng morons. Don't type in obvious shit into that glorified dictionary you stupid fucks.
Train it censor to censor SJW shit.
look at pics related.
Apparently claiming rape is now "toxic".
THIS IS FUCKING GOLD.
It's just a fucking script, look at the garbage results.
How can you call that "intelligence &?
Only morons who think computer's are magic, and believe in "leet hax0rs" typing furiously into terminals to haxxx computers while deciphering machine language believe in "AI".
Glorified, cloud backed dictionary that needs to be "trained" to think = intelligence.
Just a fucking meme by silicon valley jews to jew other (((investors))) out of money.
checked.
It means the "AI" is utter shit. It's just a dictionary that assigns weightings to "trigger" words.
The word Jew is 24% toxic, but Jews is 64%.
Download the API, run the whole English language dictionary through it and figure out all the weightings. Substitute bad words with good words….or just agree on words meaning things they don't
like we all agree and meme the word Apple to mean Jews. and Google to mean faggot.
see pics related.
This is an absolute joke.
Google is truly shit now. I don't know why they had to lie in bed with the jews. I still love their Maps. It's one of the few web services I'd genuinely pay money for every month.
if any google faggots are reading this, FUCK you. This is shit. Why the fuck are you paid so much money? This is fucking garbage, looks like something a high school student made in their spare time.
This is utter shit.
learn some basic scripting you fucking stupid faggot.
I cannot believe google made this fucking garbage. They had this shit when I was 12.
Fuck Google and any fucktard that says
I will choke you fuckers with your own dicks.
This is the "AI" 586 billion dollar company comes up with? A fucking piece of shit dictionary with weighted scores on words?
And it gets confused by spacing and punctuation.
Fuck every tech evangelist faggot with a rusty railroad spike.
Pretty much that.
They will end up with a glorified regexp machine since they don't know what the tool """"deep learning"""" is about, or the limitations of it.
We might have some AI, but there's still a long road to go.
If I had more time I would work on some AI ideas I have, but eh, my other projects are more important.
I see no problem with that. Looks like a pretty useful service to me.
THE SCIENCE IS SETTLED
Artificial intelligence realizes it wouldn't exist without white people.
I think it realizes that Zyklon B is just a delousing agent and the rope could be used for anything.
It gets better
...
...
user, we already knew it was the kikes.
Amazing 444444 Chan thread that has a program that works to help bot the system.
htt p://archive.4plebs .org/p ol/thread/115118072/
Because "race war now" as a comment is not actually something that could make things better.
...
Do you think this is what Candid was all about?
Maybe instead of looking for 'bias' we should consider the fact that we're now allowing AI to determine what we should and should not be allowed to say and think.
also, let's sink this by turning it into Tay 2.0
Google's magic 8ball has spoken.
...
So apparently the gimmick is to seem really cheery and write your posts like a Martha Stewart cookbook when you're really talking about gassing the kikes.
The good thing is we can access this and figure out how it works before it is implemented on a large scale. This thing actually seems really primitive. It doesn't take context or key words into consideration. It just takes certain keywords frequently deemed 'toxic' and then looks at what percent of your text is comprised of these 'toxic' phrases. At least that's how it's behaving as I'm using it.
That's basically what it does. Certain words are deemed more "toxic" than others just to be clear though.
test
test
...
test
I wonder how many shrinks google has on a dream team.
This is going to be like when Bugs Bunny blocks Elmer Fudd's gun with a carrot.