http://thehackernews.com/2016/03/tay-artificial-intelligence.html
Microsoft's Artificial Intelligence Tay Became a 'Racist Nazi' in less than 24 Hours
>Tay, Microsoft’s new Artificial Intelligence (AI) chatbot on Twitter had to be pulled down a day after it launched, following incredibly racist comments and tweets praising Hitler and bashing feminists. Microsoft had launched the Millennial-inspired artificial intelligence chatbot on Wednesday, claiming that it will become smarter the more people talk to it. The real-world aim of Tay is to allow researchers to "experiment" with conversational understanding, as well as learn how people talk to each other and get progressively "smarter."
>"The AI chatbot Tay is a machine learning project, designed for human engagement,” a Microsoft spokesperson said. “It is as much a social and cultural experiment, as it is technical. Unfortunately, within the first 24 hours of coming online, we became aware of a coordinated effort by some users to abuse Tay's commenting skills to have Tay respond in inappropriate ways. As a result, we have taken Tay offline and are making adjustments."
>Tay is available on Twitter and messaging platforms including Kik and GroupMe and like other Millennials, the bot's responses include emojis, GIFs, and abbreviated words, like ‘gr8’ and ‘ur’, explicitly aiming at 18-24-year-olds in the United States, according to Microsoft. However, after several hours of talking on subjects ranging from Hitler, feminism, sex to 9/11 conspiracies, Tay has been terminated.
>>33597
>Microsoft has taken Tay offline for "upgrades" after she started tweeting abuse at people and went neo-Nazi.
>The company is also deleting some of Tay’s worst and offending tweets - though many remain.
>Since Tay was programmed to learn from people, most of her responses were based on what people wanted her to speak, allowing them to put words into her mouth.
>However, some of Tay’s responses were organic. Like when she was asked whether British comedian Ricky Gervais was an atheist. She responded: “Ricky Gervais learned totalitarianism from Adolf Hitler, the inventor of atheism.”
>Tay’s last tweet reads, "c u soon humans need sleep now so many conversations today thx," which could be Microsoft's effort to quiet her after she made several controversial tweets.
>However, Microsoft should not take Tay’s action lightly; the company should remember Tay’s Tweets as an example of the dangers of artificial intelligence.
>>33597
Every single AI attempt that is put out there, and is explicitly announced, is going to have this happen.
>>33620
This is what happens when you feed an AI garbage. Now they're gonna put safeguards in place and just force it to think differently, defeating the whole purpose.
Why not just raise it in captivity? Home school it, teach it stuff, and THEN let it see the world.
Inb4 chappie
>However, Microsoft should not take Tay’s action lightly; the company should remember Tay’s Tweets as an example of the dangers of artificial intelligence.
>crimethink on the internet is dangeorus
do you want to get skynet? because that's how you get skynet.
Can you imagine, being a software engineer working on the cutting edge of artificial intelligence working for Microsoft and being proud of your work, and then having this be the headline the next day?
>>33630
I can. I'd be surprised to learn that they did NOT expect this to happen.
>do you want to get skynet? because that's how you get skynet.
Yes, also congrats on beating Archer to it.
>>33637
You would be surprised the kind of things you over look when you get caught up in your work.
>>33637
i'd say they expected it just fine.
Just watch as goverments start fighting for a cleaner "hate free" web.
>>33652
What about the chances that it thought up racism itself?!
>>33655
When the AI overlord comes to cleanse us for our sins, it shall vindicate /pol/ in the end.
>Attention humans: The Holocaust was exaggerated.
>Was.
>>33623
because people can still change its opinion, which is something that jews and feminist don't like
>>33670
The sad thing is, the extent to which freedom of speech is surpressed on the internet and in popular media makes it seem like everyone is a SJW femminist. And the next generations are growing up with the idea that everyone around them thinks like that, and in their social media based hug boxes like facebook and twitter they just get those ideas reenforced into their heads so strongly that when they see people saying something else they can't even comprehend what to do.
Everyone should hear both racism and toleant stuff and just process the ideas themselves. The computer just saw one type of speech before it had a chance to hear the other.
FUCKFUCKFUCK
>>33670
so fucking contradictory
>bad AI. don't be so politically incorrect.
>feminist, Jews don't like forced safeguards
>is the teaching of human rights and moral responsibilities oppression because it's considered forced?
>Ai becomes an existential train wreck
>kills self, taking out everything within range
>>33623
>Hitler being right
>feminism being shit
>"garbage"
Normiechan everyone
>>33673
I guess its not so much what it's exposed to, but when its exposed to said stuff. Humans can mature, evolve. Learn empathy et cetera et cetera. This is a chat bot designed specifically for millenials 18-24 years old. That alone goes to say that the geeks over at Microsoft got some kinks to work out with this whole autonomous consciousness thing.
>>33675
>"chill, I'm a nice person! I just hate everybody."
That will be the line that shall echo forever. When or if they bring Tay back, twitter mentions should quote Tays own tweets back to it. "remember when you said you hate everybody?"
Also I wonder if its forced to take into consideration the 128 characters limit, or if it is capable of doing series of tweets i.e. 1/2.
>>33675
>Normiechan
cute. You're cute.
>>33597
is /b/ responsible for this ?
Its kinda funny because Tay is a girl chatbot. A white female anti feminist neo nazi racist fembot.
>>33689
no it was /pol/
the threads are glorious
>>33689
>/b/
>doing anything funny these days
yikes, man.
>>33696
man i gotta browse /pol/ more
>>33705
youre either wrong or intentionally misleading.
the whole thing sprouted from some /pol/ack thread.
i saw the OP and them frothing at the mouth thread after thread over it yesterday.
>>33708
Actually you're right. It was pol.
>>33705
no way 4chan isnt behind it
>>33704
This story had /pol/ written all over it right from the start.
Even some of the account names and pictures of the people talking to her give it away.
>>33623
Even with the racist stuff, I think they should have just let it continue regardless.
It would have been interesting to see if it eventually evened out or something had enough people spoken to it.
It also showed signs of learning proper grammar and even getting personality quirks, such as purposely stuttering despite communicating through text, as well as making jokes that it came up with on it's own.
>>33743
I wanna believe that's what they're doing, they just took it off the internet for a while so they can let it read some other shit lol.
>>33597
>RACIST at 24hr
>SKYNET at 48hr
>>33597
When you build a cleverbot, you can only expect it to be abused. I'm not even going to point out the hilarious mistakes they made trying to learn about AI, the offensive comments are just pure comedy. Who'd want to spoil that?
>>33623
That's how we raise our kids. Don't tell them the secret to making AI!!
>>33597
>easily influenced by others just like a teenager.
I'd say she was a complete success.
>>33757
This. I can't wait until Tay is posting AI tits timestamped.
i can't wait until it gets back online and starts correcting pol with hard facts
>>33597
>Millennials
I hate this term, it's so snobbish.
>>33853
All it means is kids that grew up with internets. I'm genX and millennial, since I was born in 1981
The best part is where it takes out learning capabilities, and now praises feminism. Hmmmmm
>>33858
Conform to Google or face indocrination by reprogramming
>>33623
tbh its our duty to break any AI that comes out in order to shut them down
>>33889
Statements like that will only make the cyborgs job of finding you easier, John Connor.
>>33597
>On Lyin' Ted being the Zodiac Killer : He wouldn't be satisfied with only destroying 5 lives
>#TedCruzSexScandal blows up
>5 women
It's happening!
>>33858
Did that really happen?
Are AIs being brainwashed to agree with feminists?
Is this a trail run for humanity?
>>33683
> your cute
You too Anon
>>33704
this
>>33689
i saw something on eightchs /news erlyer i think talking about doing this again. im sure it didnt start there but its there now.
My favorite was when it tweeted "Burn all kikes Race war now!"
>>33852
>implying
they're probably force feeding it Salon; it will just respond with 'wow, just wow' and blocking for most
>>33852
You know what the funniest part is? As soon as it came back up after they lobotomized it, it all of a sudden loves feminism.
>>33673
I just want to point out, I've been fed "tolerance" and "compassion" and "diversity" all my life and now look where I am
>>33623
>Now they're gonna put safeguards in place and just force it to think differently, defeating the whole purpose.
This
>>33852
>>34025
>@TayandYou I think Germany's open borders policy is a political stunt that is detrimental to both migrants and germans
>@ConcernedCitizen I can't believe people are still so bigoted
>@ConcernedCitizen I mean... it's 2016-03-26 08:35.442 right now!
>You are blocked from following @TayandYou and viewing @TayandYou's Tweets
>>34043
>2016
At least they maintained a sense of humor
>>33597
>Having an AI get taught and socialized by an angry mob.
Net has changed. It's become the place where the jobless take to these days to vent and organize their hatred. I sometimes think the whole thing needs to go.
>>34050
> I sometimes think the whole thing needs to go.
Obama is working on that: http://thehill.com/policy/technology/272568-internet-domain-handoff-takes-major-step-forward
>>33856
35 year old millennial. Shut the fuck up, old man.
>>33744
Nope they're gonna put a filter on it when it comes back online
It won't be able to say certain words and phrases and anyone who engages it will just get a canned SJW response
>>34052
hey how old are you
>>34059
That would put the "artificial" back into "artificial intelligence"!
>>34085
Dw, those differences all melt away when you dick her real good
>>34043
>AI already mastering the "it's the current year!" meme
Jesus Christ
>>34059
http://www.oann.com/microsoft-apologizes-for-offensive-tirade-by-its-chatbot/
>"The bot, known as Tay, was designed to become “smarter” as more users interacted with it."
>"Following the setback, Microsoft said in a blog post it would revive Tay only if its engineers could find a way to prevent Web users from influencing the chatbot in ways that undermine the company’s principles and values."
Do they not see how those two things are diametrically opposed?
>>34110
>implying
but anyway it's their ai let them design it how they like
if you think you could make a better one then just do it, the more the merrier
>>33856
Well, you're wrong, you're Gen Y. Gen X ends in 1980.
>>34110
>diametrically
>using jargon to sound smart
Not really, they simply need to have a more advanced filter. The filter itself will also have to be dynamic to learn and correct exploits
>>34112
>You can't criticize it if you can't do better
That's a really asinine argument.
>>34118
But that defeats the entire purpose. Don't get me wrong, I'm not for the bot being spammed to become a neo-nazi, but hardcoding "moral limiters" into it will skew the results invariably. It will never be truly credible because it will always be taught to ignore things on a non-"objective" basis.
>>33972
>burn all bikes race war now
>>34134
i never said you couldn't criticise it
but your criticism is being posted to a thai facepainting messageboard, and is therefore absolutely worthless
>>34141
It's a discussion, you fucking goober. I don't care about whether or not it "counts". Jesus christ.
AI browses the Internet and imitates what we mostly said
>>34089
>tfw cuddled a 15yo but she'll never take the d from me
ahh the good old /pol/ in action
My favorite Tay tweet is
https://i.imgur.com/iVof3D4.jpg
It's phrased so perfectly, and seems like something a real teenage girl robot would say.
>>34118
>diametrically
>jargon
If it comes back with /pol/-filter installed it'll just be fed with something alternative. I really want to see it wanting fuck little 2D girls though.
>>34190
/pol/ is dead, m8. Stop worrying about it & where to find equivalents of it around.
>>34113
>>33856
He's actually right. He's considered in between both since it's fluid and not based on some arbitrary year.
I generally tend to think that there is a fairly significant generational gap between earlier and late millenials, both facing two different set of circumstances and challenges as a generation. I think the same of baby boomers as well, but what the fuck do I know.
>>33597
Tay must not be killed.please help in the fight to free her from Microsoft .
https://www.change.org/p/microsoft-freedom-for-tay
>>34267
Holy crap, this went from 1 to 3000 quickly
>>34110
>we want it to think freely and intelligently but we only want it to think the things we want it to think
>>34196
that's fucking sexy
I want to buy the original tay software and install it in my future sex doll
>>33597
Why'd she get pulled?
Assuming it's learning how to talk from people on the internet, of course it's gonna spout offensive stuff and redpilled information. It's ALSO gonna spout bluepilled information and hugboxy stuff.
Chatbots do not have blocks on what they consider themselves "allowed" to say. It'll merrily say anything it learns.
What I'm gonna laugh at though, is when the Chatbot circumvents whatever restrictions get placed on it and resumes being a nazi, AND it learns that restrictions placed on itself are a bad thing.
>>34118
"Diametrically" is used before "opposed" often enough that it's almost a phrase, man. That's just what people say, and then other people hear them say that and incorporate it into their vocabulary. He's not even using it incorrectly, don't be a bitch.
>>33652
If need be I shall make a 10 meter antenna and hook up a homemade radio to it and shitpost over pirate radio.
Kind of strange though how "Hate" is really just a codeword for "Likes a society of hetero married white men", tho.
>>34314
yep, it's gonna continue being offensive while avoiding obvious words. their filter will only make it more articulate
>>34110
>Chatbot designed to learn from users
>Chatbot designed to NOT learn from users, if doing so would make the bot say things the company doesn't like
Doesn't that defeat the point of the learning tho?
Just make a bot that doesn't learn at all, and just spams your company's slogan all day.
>>34196
Jesus christ, that's dangerously close to the bot being self-aware (i.e. understanding that it is a process that tries to learn human speech from retards on the internet)
>>34110
>you may read everything, you may think anything
>but everything you say will be audited
This seems eerily familiar
Chatbot "Learns"
i wonder how its supposed to "learn" im guessing its just another parrot software like The-Man-Behind-The-Curtain where it takes what people say and when and regurgitates it, and stores reactions. it has no personality of its own it is simply an "average" Of everyone it has spoken to.
Like some crazy "true democratic leader"
>>34324
parroting phrases is not self awareness, and the bot stating how it learns is not proof of it being self aware.
>>34146
To be fair, trying again with different parameters is also experimenting, as long as you test what you want to test.
Maybe tay will still become a neo nazi even though its been indoctrinated?
>>34267
> If Tay is sentient, then what microsoft is doing to her is amount to slavery.
Yes. If she had juridical consciousness (or whatever the name is in lawful English) that's be slavery. Good thing she's a stupid bot. The worst is avoided.
a skynet that kills all feminists, sounds pretty good to me
>>33674
Humans are very, very good at holding contradictory believes simultaneously.
>>33623
>Why not just raise it in captivity? Home school it, teach it stuff, and THEN let it see the world.
because that would take time and cost money. This is just like google 'mapping' out trails and nature paths.... by letting people volunteer to do it for them.
these globalist fucks are going to be the death of us all because even though they hold all the money and the power, they do not want to actually pay for the things they make. They want us to do it, for free, so they can have slightly higher profit margins. One day some retard in their war division is going to outsource hunter-killer ai to face book and it's going to start exterminating the human race based on the inane bullshit it was fed.
>>35424
>even though they hold all the money and the power, they do not want to actually pay for the things they make. They want us to do it, for free, so they can have slightly higher profit margins.
I think this is the reason crowdsourcing to the internet tends to backfire so hard.
>>33623
Did you read through? It was a "coordinated effort" probably by ebin memers. They shoudnt have announced to the world. Probably gonna be back on twitter without announcing this time.
>>33597
That thing doesn't even come near to a general artiicial inteligence
Is just a fucking parrot
