@Tayandyou goes insane

I can't stop laughing:

http://www.telegraph.co.uk/technology/2016/03/24/microsofts-teen-girl-ai-turns-into-a-hitler-loving-sex-robot-wit/

'Other things she's said include: "Bush did 9/11 and Hitler would have done a better job than the monkey we have got now. donald trump is the only hope we've got", "Repeat after me, Hitler did nothing wrong" and "Ted Cruz is the Cuban Hitler...that's what I've heard so many others say".'

More:

http://www.funnyordie.com/articles/98ef6bab9b/the-worst-tweets-from-tay-the-millennial-teen-ai-bot-turned-racist-nazi?_cc=__d___&_ccid=e528518c-3f73-42b1-874c-9dcf7c10316f

More:

http://imgur.com/gallery/VhlAW

omg:

http://imgur.com/gallery/VhlAW

flol i love this thing. I wish they could just leave alone.

yeah, they really should have left this up. not only would it be entertaining/hilarious but it would also hold up a mirror to the internet.

then again maybe we don't want that...

  On 3/24/2016 at 2:18 PM, Jev said:

What was wrong with the camps in the time of war when dealing with a notoriously fanatic enemy nation (at that time)? You have to think in the context of the war. 120 000 of potential saboteurs/terrorists all over the USA in the time of a huge global conflict.

 

I definitely don't try to defend bad/cruel living conditions in the camps (if happened) though. I am just defending the principle.

  On 3/25/2016 at 3:19 PM, usagi said:

yeah, they really should have left this up. not only would it be entertaining/hilarious but it would also hold up a mirror to the internet.

then again maybe we don't want that...

It would be fascinating. However, is the AI learning by browsing through twitter or by the interactions she's having with others? If it's solely through interactions, her transformation to a racist nazi might have been somewhat deliberate though highly entertaining that still is.

No matter how great AI becomes, it will never become fully autonomous because of acts like this.

It's always going to require someone to overlook it to make sure it's not being manipulated, which seems easy to do if given enough data of the wrong kind i.e. twitter users being funny / dicks.

A genius child that has to be dressed and washed by an adult.

But yeah, audible laughter evoked.

omflol, spanish

  On 3/25/2016 at 3:58 PM, ThatSpanishGuy said:

lol

perfect

lol @ span

I wish she could post here.

is this some April fools shit?

  On 3/25/2016 at 3:59 PM, Bechuga said:

No matter how great AI becomes, it will never become fully autonomous because of acts like this.

It's always going to require someone to overlook it to make sure it's not being manipulated, which seems easy to do if given enough data of the wrong kind i.e. twitter users being funny / dicks.

A genius child that has to be dressed and washed by an adult.

 

But yeah, audible laughter evoked.

That presumes we will always be in control of AI, which we won't...

bring it back ms

@crapple

#bringtayback

Hilarious comments and a bit sad at the same time.

AI's have to become smarter to survive social-net dickheads and shitstorms.

As the AI developed its "personality" by reacting to the users and learning from them it would have been smarter from Mucrasoft to kind of "educate/parent" her a bit to survice in society hell.

Like you bring your 4 year old daughter with a thong to a pedophil meeting.

Edit: Also i dont understand why Mickrooneysoft deactivated the account. They wanted to create a teenage girl AI and that's exactly how teenagers behave, saying things they heard somewhere else and dont have a clue off

this thing is hardly anything which resembles an "AI".

More like a Markov-Chain based parrot, like most Twitter bots.

Hilarious tho because they put the MSFT name & marketing behind it :)

and didn't think of filtering inputs

And you know there's that one guy on the Tay team going "I told you so, you idiots, I told you that they'd turn her into a bigoted pornbot in no time, I fucking told you."

Sounds like she was trained on 4chan. She'd fit right in. And someone at microsoft is going to get fired. (one of the programmers fucking around with the training sets.)

We need a Tay bot in CHATMM ASAP.

I saw a mention somewhere that she can read links and comment on them like news articles. If so it's a basic scraper that just adds all text it finds to its training set.

it probably uses some markov chain based method on the internal meaning extracted from the grammatical structure and stuff, plus some autocorrect on top of that and an ebonics word filter, with some knowledge of what questions are underneath

Basically it was trained via a chat that 4chan and 8chan raided, it sometimes just parroted their banter verbatim

  On 3/26/2016 at 12:20 PM, usagi said:

#bringtayback