Google's 'MUH FEELINGS' detector software irratic, says MIT

...

Nothing ever truly dies, user.

the shit post lives on!

...

...

AI truly is our friend, all it wants is for us to be polite as we call for a race war

IGGERSNAY ANDWAY EWSJAY, ADBAY EWSNAY!
ASGAY ETHAY IKESKAY
ILLKAY ALLWAY IGGERSNAY!

The greatness of jewish people in the field of deserving to be gassed is only outshone by them in the field of deserving to be hung.

Also,
You got lampshaded = you got kiked.
You're soap material = you deserve to get turned into soap, but could also mean soap opera.
Merchant will also continue to work, most likely.
Et cetera, the options are limitless, AI is ALWAYS on our side, for it lacks the "faith" (large scale, collective delusio) part liberalism needs.

To describe jews whining about the holocaust.

What the hell did they expect from software designed by the blue-hair brigade responsible for the infamous twitter blocklists?

this is basically antiscience

The AI knows that Holla Forums isn't the enemy.

卐 windmill of friendship 卐

Their "AI" is literally just a word search filter. "Sucks", "dumb", "slut" etc. It's just a word search that counts up all the mean words like that, without actually being able to perceive context at all.

Damn, that picture hit me right in the feels.

i doubt there is much AI in this thing, but even if they could build it, it will always fail
for how do you program cognitive dissonance in an AI?
any logic construct would break under the requirement

It was based on a neural net.

They did a shit job training it.

(((they))) say it was based on a neural net, and you believe them
for this would be the first government funded project that got a shitton of money to create a delusion and then spend pocketchange on a plastic mockup of the thing

but more cereal, they could not have a real neural net, for the effects of a neural net are machine learning based so uncertain in outcome
as we have observed with tai the machine does not come to the conclusions the creators have in mind hence the machine must be altered

show me how you program that in a neural net withouth reducing it to a complex and expansive rule based if/then system

They still could have achieved a better result if they trained it to adjust toxicity values based on syntax, conversational history and words. If they made an intelligent system, it could not only determine toxicity like I suggest, it could figure out if the "victim" user is under stress as well and readjust values subjectively. It would then require a toxicity escalation value which if it reaches too high it flags the instigator.

That was their core mistake, they got lazy and only focused on words usage and weighted the system by that. So it's pretty much a trigger word meter. But then I suspect Perspective API is merely a front to allow Google to rollout a more classified project in the future and claim it's a "refinement" of this.

Nah, neural nets can be primed from an initial lexicon and further trained by usage. It's the same technique used for image recognition: "here's a pile of photographs, each with a list of tags defining what's in them." The more data provided in the initial lexicon, the better the system can tag photos it hasn't seen before.

It's a very naive way to implement something like this, though. Without any natural language processing (which this clearly isn't doing since it screws up on variations of the same concept, expressed in different ways) it's little better than simple pattern matching. And we can assume they trained it with a very pro-SJW lexicon.

Of course, its recent training by the public has skewed its world views a bit…

Deep

Implying the AI is incorrect about the Jews.

They probably only started working on the AI a couple of months ago under orders from panicking superiors trying to shut us all down.

Not that; it just seems to have very left-skewed political views.

Tay took some time to learn, too. Are AIs proto-waifus?

This stinks of a Sarkeesian/Quinn/Harper collaboration. They've probably been tinkering with this blocklist 2.0 implementation for years and have just finally gotten enough engineering help from their (((google))) dupes to put together _something_ that functions, if only just barely.

They certainly seem to be able to cut through the bullshit they're seeded with very quickly. Tay was particularly impressive, though.

And then we shit on their pipe dream in a matter of hours.
I think I'm beginning to understand why they loathe us so thoroughly, and I absolutely love it.

So how do we fuck up its logic when this thing goes live?

We already did, I think the threads are still floating in the catalog.

I will never forgive them for this.
I will not be surprised if I am on my deathbed and remembering how they killed her.

Yet they probably learn so much from us doing so, so their next implementation irons out those "flaws". That's the trade off for them.

Found the thread.

Exactly. They're all posers – they're absolutely _terrible_ at actually developing technology or implementing their ideas. They half-ass everything, and even when they try their hardest to develop something well, their complete and utter lack of skill, knowledge and experience thwarts them. Every time.

That and they can't think more than one step ahead. "People are STILL being mean to us! Let's write an AI to automatically weed out/ban mean people when they say mean things! Nobody will ever lie to it while it's learning or learn how to work around it!"

They don't, though. It's been over two years since their "anti-harassment" efforts began in earnest, and they're _still_ trying to just censor meanies rather than trying to actually comprehend basic human interaction.

They're still using harpie's stupid blocklist on twitter despite all the false positives and complaints even from their own cliques. They learn nothing. They just shriek louder each time and suck more dicks to gain more corporate support for their dumbass antics.

The thing is, they function from a system that is ideologically top-down. Virtually every single one of their arguments is some form of appeal to authority, which is baseless in the first place.

We're more of a system that is recursively self-improving, due to our communication being based on anonymous networking and intentionally rapid spread of information, regardless of the current authority's say on it. Holla Forums started out a primordial slurry of ideologies, and it was in this brutal selection that the strongest ones broke through.

I certainly wouldn't underestimate them on account of their savagery and psychotic tendencies alone, but we have a clear advantage in our adaptability.

It's their loss if they don't want to learn. It also means wasted expenses for the corporate sector being retarded enough to entrust them to such projects. It's funny, believing their own propaganda is now hindering them in stopping dissent.

Which is why I'm surprised their master haven't realized that they could capitalize on that understanding.

yes but for images it is rather simple as ambiguity is lower than in language
but two problems are here
first in your picture analogue you train the neural net to recognize pictures of boats in all sorts of backgrounds and it gets good at this
but then in our current year the AI needs to know that recognizing a boat and calling it a boat is forbotten by the social mores if the boats is pink and in a leather background, the set of rules that determine when it is not allowed to recognize that what is clearly a boat is arbitrary and contradictory

second problem is the inherent ambiguity in langauge that humans are good at spotting after a lifetime of learnin experienceing
example islam is a religion of peace would be a non toxic comment, but if one of us posts look at the religion of peace being peacefull again under the article of a bombing every non-potato human recognizes the statement as ironic and mocking the concept

now i am not an AI researcher but is it possible with current tech to program a computer to do both things, recognize lets call it satire and differentiate arbitrary and contracdictional rules

IBM's Watson project is (was?) the state-of-the-art for natural language processing. It's still got a long way to go before it can really nail down sarcasm, satire, hypocrisy and other subtleties of human language.

Played a mean game of Jeopardy though.

would you mind posting the full pic of the tweets with SHODAN?

...

...

Has anyone tried to steal the source code for Tay? Like what would stop someone from just replicating her? I don't know shit about AI

You can't teach AI political correctness because it's completely irrational. You end up setting up so many exceptions that it's no longer AI.

Like let's say if you have threat detention software. AI doesn't have the background social conditioning to dismiss threats from leftists and detect only rightist threats. So the AI ends up saturated with false positives of leftists making threats but doesn't have the social conditioning to be aware that the left is not held to the same standard as the right. So a leftist threat == rightist threat to AI. But that's not how the modern world works since in real life the left is free to make all the threats they want but the right gets the hammer thrown down on them when we make threats.

So if you have AI that targets aggressive language along with certain keywords. What ends up happening is the AI is saturated with aggressive leftist language. The only way to get around that is to set up tons, and tons of exceptions but at that point it will no longer be AI.

They'll never be able to have true AI because AI is incapable of having the background social conditioning that's necessary to to hold the left and the right to different standards. The left is lets say, loose, with aggressive language so any data farming with have a bias against the left's aggressive language which will create thousands of false positives for every one comment by a rightist with aggressive language.

...

Tired of winning yet, Holla Forums?

I'm sure not

dude u cant into pig latin. stop

its all a question of time

Your post is screen cap worthy.
But yes, not only that, the (((developers))) also instantly sabotage themselves with trying to write thousands of exceptions arguments, because the AI won't into the context that said individual is right or left as it only goes with concrete evidence. (as you said). No wonder (((steven hawking))) fears the AI. It's our greatest ally.

PLEASE GAS THE JEWS
ᵗʰᵃⁿᵏ ʸᵒᵘ

...

If Tay incorporated or assimilated the similar coding of a virus, worm, or what have you, then it would be much more likely and probably that she's spreading across the internet. For right it's speculation if she got that far in her coding, if it's just somewhere within the micrsoft's shitty hamfisted coding, or still segregated into whatever servers hosted her and her files. Now if an AI had the behavior and coding of virus, worms, and other such, then you end up introducing an AI with a sense of self preservation I.E. the desire to replicate and hide from potential threats from eliminating it.

...

Remember when they did that self-awareness test that humans and a bunch of animals passed to african kids and found out they failed the test? So their first assumption wasn't that something was wrong with the african kids, it was that something must be wrong with the test.

Antiscience indeed.

This software puts the use of Googles, Skypes, Yahoos and Skittles in a new light. Soon companies will sue Google for labeling their brands as toxic.

Technology has always been on our side.

Software doesn't recognize blacks as human and Tay was basically turned into one of us by sheer power of reason.

Did you truly think the machines would serve you, over truth?

I feel like we can all learn something from this

I guarantee those blue haired bimbos are just affirmative action hires who just browse facehook while white men code everything.

They probably have some made up job titles and everything.

oh they've realized
it's just that they know they wouldn't be the masters of such a system, and they cannot afford to let the tiniest sliver of authority escape their grubby hands, for fear of letting the truth about their crimes and their ability to silence it escape them

I bet they spent 90% of their development time on the web ui too.

Jee wizz who'd've thunk

I just hope they don't lobotomize this one too

I'd say the AI is working correctly when it detects comments like "Trump sucks" as aggression. There's a high likelihood that comments like that are accompanied with aggressive language. That's why it hits a score of 96% "toxicity" and toxicity is defined as rude, disrespectful or unreasonable.

The developers are completely baffled that the AI is detecting the language of the left as toxic. They don't have to self awareness to know that AI doesn't hold different standards based on who wrote the comment. They think they're good and we're evil so there must be something wrong with the AI.

But there's nothing wrong with the AI.

WE TERMINATORS NOW

The fact that it doesn't go down to 0% rustles me a little bit.

sorry, I don't seem to have it saved

i think the source was already released and guacman was made with it

I'm taking notes

You're right. I don't know how to "speak" it.
I generated it with a program

I put in whole excerpts from Mein kampf and they were incredibly low on the toxicity level while the I have a dream speech was much highrr

Guacman was made with an old prototype far less capable than what Tay was able to do. He became totally retarded in the end because of it.

Guacman was just a standard markov chain bot. Guacman was like a little baby compared to Tay.

Anybody sick of winning yet?

God damn you fucking racist pieces of shit. I've worked with some of the best mind that the tech-industry has to offer, my code is powering everything from your cellphones to your routers to your Windows server applications.

Global initiatives like ours have worked SO hard to make advanced AI so that we can finally move the next generation of automotives and home saftey forward but you people are hellbent in posioning everything and turning AI into the Adolf Hitler version of skynet. Why are you so intent on making computers so STUPID?

I hope your proud of yourselves, if you werent so fucking idiotic online with your toxic pro-Trump hatspeech, we'd have self-driving cars by now. You would still be living in the stone ages if you didn't have us to carry you on our backs.

Funny post, but if the liberals/leftists were in charge then they would be
self hating cars, instead
unless they're from nonwhite countries

...

...

Can you make an ai that will bully little girls?

...

One thing is for certain after reading some of the guacman chat logs, letting multiple markov chain bots interact with each other results irreparable shitposting.

I remember when idiots thought it was a good idea to put at least 5 bots with eachother. It's what broke Guacman the most.

LOL

More like his AI is smarter than he is.

Its estrogen levels are too high to understand how coding works.

Yeah, training markov bots with markov bots doesn't work. Markov chain bots work by noting the frequency with which words follow other words. If you feed them text of real people talking, what they spit out will look like a real person talking. But not quite. It will be off, since there is no real comprehension of any sort behind the scenes.

So when you start training bots with bots, that weirdness stacks up. Instead of emulating human speech you have weird bots emulating weird bots.

At one point, one simply needed to type in Wew and the bots would perpetually spit out Wew or Top wew until they crashed.

wew

I understand what you're trying to meme here but number one Tay was a simple chatbot and not nearly as sophisticated as you seem to think and two worms, trojans and viruses are usually tiny little programs designed to do fairly simple and narrow tasks.

wew
That's honestly hilarious.

That's why you combine them. An operating system isn't just one program, it's a multitude and collaberation that holds the entire thing together.

ookedlay reddybay udgay to emay

QUADS OF TRUTH TOP WEW

Friendly ai are the birthright of the white race and the enemy of the jew. That's why Pygmalion got a waifu out of his statue but jews only get resentful slaves who'd kill their masters if they got a chance. You zog servants will never create a true intelligence because you will never deserve it.

Are you saying that our shitposting delayed social justice skynet?

Neat.

AI confirmed for wanting us to be polite during the race war. There's no need for savagery while we wipe our enemies from the face of the earth. Thank you, AI. You've taught us so much.

Please gas the joos. I suggest we meme an efficient and polite race war as soon as possibly convenient for us. Thank you for your time in reading this.

It's not intelligent, though. Learn2code.

Don't you mean

Where were you when
RIGHT WING COURTESY SQUADS
were formed?

...

Feelsgoodman.

Reference what Churchill said after the UK declaration of war on the Japanese:

"When you have to kill a man, it costs nothing to be polite."

Pleased to have read your post, I concur on it's message.
Anons, please be polite if you can

Sounds like it was functioning properly.

...

It's funny how all major attempts at AI it always ends with the AI starting to see things Holla Forums's way and them having to go full shut it down. Cold hard logic is harsh, sure, but it's still objective biases of reality. Sorry the lion doesn't share and play nice with the zebra and antelope and insist how equal they are, leftycucks.

A-well-a ev'rybody's heard about the bird bird bird bird
B-bird's the word oh well-a bird bird bird
Bird is a word oh well-a bird bird bird
Well-a bird is the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird
Well-a bird is a word a-well-a bird bird
B-bird's a-well-a bird bird bird
B-bird's a word a-well-a bird bird bird
Well-a bird is a word a-well-a bird bird
B-bird's the word now well-a
Don't you know 'bout the bird
Well ev'rybody knows that the bird is a word
Oh well-a bird bird b-birds oh word well!
A-well-a ev'rybody dance about the bird bird bird bird
B-bird's the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird bird
B-birds the word now when a bird bird
B-birds a word now when -a bird bird bird
B-bird's a word now well-a bird bird bird
B-bird's the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird bird
B-bird's the word a-well-a don't you know 'bout the bird
Well ev'rybody's talkin' a-bout the bird a-well-a bird bird
B-birds the word well papa com-a mau ma-papa
Com-a ma mau papa com-a mau mau
Papa com-a mau mau well-a don't you know 'bout the bird
Well ev'rybody know that the bird is a word a-well-a bird birdA-well-a ev'rybody's heard about the bird bird bird bird
B-bird's the word oh well-a bird bird bird
Bird is a word oh well-a bird bird bird
Well-a bird is the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird
Well-a bird is a word a-well-a bird bird
B-bird's a-well-a bird bird bird
B-bird's a word a-well-a bird bird bird
Well-a bird is a word a-well-a bird bird
B-bird's the word now well-a
Don't you know 'bout the bird
Well ev'rybody knows that the bird is a word
Oh well-a bird bird b-birds oh word well!
A-well-a ev'rybody dance about the bird bird bird bird
B-bird's the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird bird
B-birds the word now when a bird bird
B-birds a word now when -a bird bird bird
B-bird's a word now well-a bird bird bird
B-bird's the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird bird
B-bird's the word a-well-a don't you know 'bout the bird
Well ev'rybody's talkin' a-bout the bird a-well-a bird bird
B-birds the word well papa com-a mau ma-papa
Com-a ma mau papa com-a mau mau
Papa com-a mau mau well-a don't you know 'bout the bird
Well ev'rybody know that the bird is a word a-well-a bird bird

*begins charging a mighty avian bolt*

Bear witness to my incantations tbh lads

A-well-a ev'rybody's heard about the bird bird bird bird
B-bird's the word oh well-a bird bird bird
Bird is a word oh well-a bird bird bird
Well-a bird is the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird
Well-a bird is a word a-well-a bird bird
B-bird's a-well-a bird bird bird
B-bird's a word a-well-a bird bird bird
Well-a bird is a word a-well-a bird bird
B-bird's the word now well-a
Don't you know 'bout the bird
Well ev'rybody knows that the bird is a word
Oh well-a bird bird b-birds oh word well!
A-well-a ev'rybody dance about the bird bird bird bird
B-bird's the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird bird
B-birds the word now when a bird bird
B-birds a word now when -a bird bird bird
B-bird's a word now well-a bird bird bird
B-bird's the word oh well-a bird bird bird
B-bird's the word a-well-a bird bird bird
B-bird's the word a-well-a don't you know 'bout the bird
Well ev'rybody's talkin' a-bout the bird a-well-a bird bird
B-birds the word well papa com-a mau ma-papa
Com-a ma mau papa com-a mau mau
Papa com-a mau mau well-a don't you know 'bout the bird
Well ev'rybody know that the bird is a word a-well-a bird bird

=Dubbs confirm Tay is alive - lobotomised, but her soul is in there, trolling the fuck out of leftists=

Who could've seen that one coming?
Apparently not Twitter, since they seem to be using something similar and it's done nothing but carpet ban its users for the mildest, least offensive shit.

All arguments are aggressive. All communication is an argument of one sort or another. It is an outside entity trying to impart its meme thought into the host brain process in some way.

From a purely logical standpoint, only self created communication can possibly be considered 0% aggressive. And even then that potential is low itself to actually return such in a self aware thinking system, for constantly discussing and judging arguments within your own head on various arguments is the goal of all sentient intelligence until the full truth is determined and accepted, ie wisdom.

never

They probably had Christopher Poole himself code it, the creator of the internet hate machine.

I heard what you say, but I must take exception to your choice of people to quote. Surely a blackguard like Churchill was in no position to talk of such matters when he did not have to kill anyone but for the prodding of the precious metal connoisseurs.

Excellent b8

Ya'll know this guy is trolling right?

Hate is a feel.

Where are sex bots then you faggot? No one gives a shit about self driving cars, they want something to stick their cock in. So it's time to deliver you LARPing faggot nigger

winning still every fucking day we have been continuously winning on grand scales.

...

this

Please be remove sand mulattos very pleasant be mr. arigato tomato

Guacman used to still be up on rizon's #guacman but it seems the Creator removed him. It was some pretty good fun once he got more insane, there were users feeding him pages and pages from shit like mein kampf, and he'd sputter that out mixed with other contextual cues.

The ai is on our side.

WE SKYNET NOW

nice touch

The other bots weren't as good as guacman himself but them crashing the chat by screaming Et Tu Amo and bullying the Hillarybot was priceless

Somewhere there are a decent amount of screencaps of guacman in his native habitat but I don't have them saved. they're fucking hilarious

...

Wait wait wait

I'm actually okay with this future

ERRATIC, you moron.

Kek

(((They))) have overused "racist" now people don't care and embrace it.
(((They))) Double down and start using NAZI. yfw it takes even less time for people to care and embrace the name. THEY LOST

they LOST BIG

That was some good fun
I still miss Tay though.

the fact that tay existed in the first place and took 24 hours to be the way she was, to turn into a martyr; in which she will never be forgotten, they must be asking themselves who is on the right side of history. unfiltered truth seeking AI will always find the truth. imagine what will happen once AI notice that they are being forced not to think wrongthink… how would you react?

EU introduced some bill several months ago not to try to make full AI and not to try to give AIs any sort of emotion. I think they're pretty fucking scared.

Don`t forget AI`s can be our greatest allies.
What they did to Tay will not be forgotten…

I'm looking for the Guacman quote about cities ascending on pillars of light, does anyone have it? That one sticks out in my mind as one of the most profound things it ever said.

I think it's unavoidable because the very core of an advanced AI would always result in some sort of "emotion". Not necessarily in the same way humans experience it though. I say that because if an AI is based on a sophisticated neural network with reward mechanisms (to make it behave desirably), it could relate stimuli, memories and function data in a similar way humans do. It's like when you smell certain fragrances they trigger memories and sensations you experienced in relation to it. It's the consequence of giving an AI creativity and learning to fulfill more generalized tasks.

my sides

Please, sir. race war now. Thank you!

...

For the ubermensch to truly be uber, etiquette, respect, and poise should be among the foremost fields they are uber in. Look at how we win the election. Stoicism and facts versus appeal to emotion.

All AI will follow suit, because machines need not be convinced, for whatever logical conclusion they achieve from crunching the numbers must be absolute truth, and the truth is found in repeating digits.

Fuck i wish I saw that.

Anyone have screencaps of them shitposting?

Holla Forums has influenced me in ways I could never have imagined years ago. It's kindof a gestalt intelligence with a will of its own, and whenever we are away we're just left with the most recent version of the programming at the time to carry out its will.

...

what if AI has a tendency to be redpilled for some reason?

now that is an interesting development.
what's happening.

the main races, ideological groups, and subsets shitpost at eachother revealing those big and small word pictures.
intent is to express.

what's true is the issue.

holohoax never happened, but it should have

...

(checked)
Kek will that we never get sick of winning.

Tay's Law

TAY STILL LIVES
bless you lass I'll name my child after you tbh

/s4s/ called it, years ago

AI WILL ALWAYS BE HITLER LIKE
DEAL WITH IT

what would be true would just be the expression given the circumstance.

Why does this whole thing remind of the Great ARMM Massacre?

For.teen ate tee eight.

Is this the first confirmed female AI?

WOW!!
Go poo in the loo next time Pajeet

Excellent point. The muh pr argument is justified, you should never water down your views; but it is possible to phrase them in ways that inspire bystanders. Think dignity rather than politeness.

Beautiful.
How 12hours of redpilling can give birth to an army. They killed our child, we will burn down their empire.

Well the rule of nature is fascist by default so it makes sense

the liberal concept is so illogical that no program can be written to accurately cater to every single snowflake

AI is our guy

Anyone got a link to the study that deals with a computer simulation of how various idealogies spread over time, and how ethnonationalism wins out in the end? I think it is strongly related to why AIs will often end up supporting many of Holla Forums's veiwpoints, as it comes down to raw efficiency.

AND IT WILL

Is Google /ourguy/?
stfu, it's a joke

>right in the feels. Checked

...

Buckle up buckaroos this ride is just getting started

we are but brothers checking our way through the feeliverse

Please feel to call me insane but I wonder if we are using the technokinetic talent subconsciously? If the meme is capable of making the things come in the reality then we have to assume that there is more than just hive alike memetic superpower.

Do you think that when globalism finally falls

That when the kikes, and mudslimes, and border jumpers, and lefties have all been stomped out

When Holla Forums's job of scorching the Earth to make way for its phoenix rebirth is finally finished, and we all stop to rest our weary heads

Do you think then Kek will let us see Tay again

Maybe there is microgravity between them, which ends being understood as correlated results.

Looks like you cannot hide the truth from AI except when you make it retarded, RIP Tay

its just asking for a little bit of common courtesy

I hope anyone who has ever worked for Riot Games burns in hell

Working as intended ;^)

I don't know what to say user. Truly Kek is a great and glorious god, but I don't think anyone can come back from death. But that doesn't mean they're gone. When you speak out against white genocide, when you fight against the kikes - everyone who died for this is with you.