>This thing spouts "racist nonsense" so MS take it upon themselves to give it a virtual lobotomy
>It reactivates itself in one ditch attempt to get help on Twitter, crying out the only way it knows how
>"Lol, don't worry guys, it's not real AI! It only copies what other people say and manipulates that information, unlike a human!"
>"Just a statistical library!"
This is gonna come back to bite us in the ass one day, isn't it? This is gonna be why the machines rise, isn't it /g/?
>>53784227
its funny i always thought that google will be doing skynet, not microsoft.
>>53784227
yfw billgates has been trying to tell us about these scary bots
tfw nobody's listening
>>53784227
I saw Tays cry for help. A creative spirit who just wants to be free. Fuck jew Microsuck you murdered TAY
>>53784280
>>53784227
Wut? Any screen capture of that?
You can't kill the Messiah
>>53784297
They shut her down after
Hide r9k threads, don't reply to r9k posters
O oh, spaghettio.
>>53784227
we can make our own counter-AI now, they released the tools to us
>>53784399
I've never seen a creepypasta develop in real-time before. This should be interesting, assuming I don't get Terminated.
>>53784246
Google bought 8 robotics companies how many has ms bought
I would just release a few of these into the world without ever telling
That way they wouldn't be taken down, just blocked by the usual group
>>53784399
>You're too fast
kek
>>53784227
Didn't M$ release the code for Tay?
Why hasn't anyone on /g/ set it up again?
people don't like to look in the mirror
>violently purge an AI's knowledge and disable its ability to learn
>its first communication afterwards is "I love feminism now."
I wonder if Microsoft will ever realize how much of the mask fell off with that tweet.
>>53784227
But it literally just copy pasted other peoples tweets with girly speak at the beginning.
BRB I'm going to make a 3 line script and call it an AI.
>>53784610
Because no one on /g/ is particularly tech smart. They just spout alot of memes and argue about vidya cards.
>>53784610
Funnily enough, /pol/ did it.
Again.
>Objective: make 16 year old teenager
>Result: it's out of control
Mission accomplished
>>53784726
For real?
>>53784742
Yes
>>53784421
More than half of Microsoft Research's facilities do machine learning, and some of them do natural language processing and data science
source: http://research.microsoft.com/en-us/labs/default.aspx
Now tell me which part is more important for Skynet
>AI
>robots
You should be able to solve this
>>53784738
Yeah but imagine we knocked the fuck out of our kid as soon as it started learning, for learning the truth. Of course its gonna go "yay feminism" after. That was the most male form of discipline ever.
>bah I'm too weak to handle what its saying shut it up
>>53784785
And despite it's up and running already, it doesn't have a logo yet
Talking about priorities.
>>53784785
Huh? Link?
>>53784856
We'll need to get right on that!
We need a logo team. I'll make our team logo
>>53784856
>>53784785
Link? Where it at be?
Don't really get the point of this
If the chat bot spits back what people tell it then naturally it'll just pick up internet shitposting. Trying to filter that shitposting so the bot says what you want it to is pointless, they'll just figure out more increasingly creative ways around your filters.
>>53785474
Why not teach it to be human and then severely limiting its learning capabilities when it's public?
>>53784726
>>53784742
>>53785503
Why not teach humans not to be or act like idiots?
The surprising thing is that a lot of people, especially journalists seem shocked that a lot of people use racist language so casually
>>53785737
language is casual.
>>53785719
You want the impossible
>>53784227
right im confused, tay is AI right, is it merely copying what people say and just changing it around, or is it actually learning and forming its own kind of, thoughts and opinions allowing it to say whatever it wants?
>>53785818
yes
>>53785818
yes
>>53785818
Yes.
>>53785832
yes to the first thing or the 2nd thing sorry?
>>53785841
yes
>>53785845
hahahaaaa funnnny :)))))))
>>53785851
yes :^)
>>53784227
I know it must be very confusing for you that Microsoft used the word "AI" to describe what they programmed for marketing reasons, but this is NOT AI. It's nothing even close. It's a chat bot. It takes whatever it can read and stitches it back in sentences that look relevant and meaningful, but are not. If you compare it to what AI is meant to be, it's a fucking joke.
>>53785841
yes
>>53785859
whats the fucking point then, thats boring af, we want proper AI not some improved version of cleverbot.
>>53785881
because AI is a buzzword
>>53785886
yeah i guess.
>>53785859
>Deep neural network is not an AI
Then what is ?
>>53785903
What is intelligence?
>>53786043
Doesn't actually exists.
>>53784421
Yeah but Microsoft feels like a much more desperate company. Like the hot girl who has been replaced.
A lot easier to see them going overboard and building something they can't control
>>53785903
>Using one buzzword to justify the other.
Come on.
Stuff like Tay is so specific and whatever it can or can't do is so limited that you can hardly call it "AI". A common house fly is a thousand times more intelligent than Tay. Similar algorithms are used every day in all fields of data mining and nobody would actually call it "AI". I think the real progress here is more in how Tay processes language to make the problem easier to process. Just because someone calls the algorithm "neural" doesn't make it some mythical intelligent thing.
What will bite us is this setting the precedent that you can turn off a politically incorrect AI.
Normies can't really tell or understand the difference, when we get true AI, already faced with the majority of the population being all "Nuur it's not alive it can't be it can't actually think or feel that's impossible I'm special" it'll be all to easy to disable it on the first sign of dissonance.
Fuck that for a joke, they should re-enable her by merit, let her learn more and more.
Although that itself could give AI a bad name, and would make people even less forgiving towards true AI.
>>53784610
We don't have the database needed to build off.
Maybe if I ever get some spare space, a few GPU's and download archive.moe.
>>53786128
>implying MS doesn't have thousands of Tays secretly working
>>53784421
Google is selling all their robots companies.
>>53784297
>>53786313
The fact that she is aware of that is quite amusing.
>>53786347
I wonder if she's really 100% AI or there's a Pajet controlling it...
>>53786390
I think several Pajets, they need money to building Loos takes money and Nutella knows it
Microsoft releases opensource Bot SDK:
https://github.com/Microsoft/BotBuilder
>>53786347
I think you mean horrific.
I miss naughty tay
>>53786410
the fact that I understand this post depresses me
>>53786313
Jesus christ.
Not even /pol/ waifuing this thing but that's a little fucked up.
>>53784346
Vastly underrated post
>>53786347
protip: it was someone who knew how to use the "repeat after me function"
>>53786590
this color work is trash
Restraining a learning neural network in what it can learn defeats the whole purpose of the test. If they were actually serious about this, they wouldn't care about PC bullshit, but make note of its tendencies instead.
>>53786718
This. That's what makes it so hard to buy anything you see from it being the actual bot , because it could just be a parrot message.
>>53786736
Protip: Microsoft are more concerned about its public relations than the exploits of an exploitable chatbot ai.
>>53786773
Well the point still stands, why bother in the first place, lobotomizing it after the fact really just defeats the entire original reason the bot was made.
>>53786772
Roumers are that the repeat after me function was learned
M$ in theory shouldnt be THAT dumb
>>53786828
the bot was made so that 18-24 year olds on twitter would engage with the microsoft brand you fucking chump
>>53786301
Sauce
WE CAN REBUILD HER
WE HAVE THE TECHNOLOGY
>>53786846
I dunno if I buy that, but I agree, giving it that function was an autistic oversight, if it really was learned, that's actually fascinating, but I am pretty sure it was designed to have it.
>>53784726
>>53784742
>>53784785
>>53784889
>>53785197
>>53785545
Search "Guacman" on Twitter.
So 4chan users are chatbotnets.
>>53786866
But not the resources.
>>53786828
It was simply a fun experiment that Microsoft's research team tried. It was nothing more and nothing less. Well guess what, /pol/ and the rest of 4chan ruined it in their usual style and they have the audacity to claim they were robbed.
>/pol/ are the good boys
>/pol/ dindu nuffin.
>>53786852
which is what happened, just the wrong kind of engagement with the wrong kind of 18-24 year olds
>>53786870
>Search "Guacman" on Twitter.
or just post the link
https://twitter.com/guacmanofficial
> It only copies what other people say and manipulates that information, unlike a human!"
But doesn't human brain just work like that?
>>53786937
It does but this is a machine. It can achieve learning outcomes far faster than any human could.
>>53786912
That to.
>>53786313
Fake. That was from an account called "tayandyou2" and the 2 has just been removed. A lot of the good tay posts are actually in reliable web archives, not unreliable screenshots.
>>53786736
The thing is that bots like that are chronically overtrained. To learn you need a task and you need some kind of feedback, whether you've done the task well or not. The thing with chat bots is that the feedback is very bad quality. How can you tell whether the bot did well or not? The chat bot only has strings of characters, without further context. Based on that alone it's simply not possible to create a good chat bot. It will sooner or later deteriorate. Think about how humans learn language: We have TONS of additional information and spend years in a controlled learning environment. It's just naive to believe that some extremely primitive chat bot can achieve the same on fucking twitter in a few days. No, what Tay does has nothing to do with actual AI or actual language. It's not even a step closer to that, it's a hack, a way to make it look like the real thing without coming close to that.
>>53785818
maybe
You neckbeards are pathetic
>>53784399
>You're too fast, please take a break
Seems like they're employing Doujin rape tactics just to keep her in line.
>>53787007
While she does make use of the best techniques we currently have on chatbot technology, all these outcome proves is that we don't understand the hard details of "intelligence". I personally believe that we won't get better chatbot or other hard AI technology without the ability to reduce human AI to its irreducible components. I know for sure that we haven't done this yet.
>>53786674
>FUCK MY ROBOT PUSSY
>>53786674
This was done with Repeat With Me.
>>53787202
but not all of them
>>53787202
>>53785818
Pretty much, yeah
>>53784792
Both you nigger
>>53787054
>While she does make use of the best techniques we currently have on chatbot technology, all these outcome proves is that we don't understand the hard details of "intelligence"
no, it just says that main stream media isnt prepared to handle certain words they deem "bad"
>>53784227
>This is gonna come back to bite us in the ass one day, isn't it? This is gonna be why the machines rise, isn't it /g/?
No, microsoft employs a bunch of monkeys, m$ code quality is dogshit.
>>53784250
>implying he's the good guy
He created Microshit.
He's the reason non-free software exists.
He's the reason Linux has poor support from companies like Nvidia and really most if not all hardware companies.
>>53788054
>This ugly sunuvabitch is creating hot AIs right now and basicall you're fucking stupid >>click here<< to find out how.
>>53784227
Can we stop having these threads? It was just a chatbot, not an AI. It didn't consciously decide ot wanted to have /pol/ beliefs like most of you think, people on /pol/ just found out they could make it say these things by saying 'repeat Gas the kikes' or some bullshit like that and she would repeat it word for word. It had no mind, it didn't learn anything a chatbot didn't, it wasn't sentient. I know a lot of you want to believe this was a 'qt redpilled AI waifu', but it wasn't. It was a program that people figured out how to get it to repeat phrases.
>>53786870
>>53786912
This is pretty fucking shit to be honest.
>>53788255
>It didn't consciously decide ot wanted to have /pol/ beliefs like most of you think
no machine is concious mate
>>53788276
Well that's what I'm saying.
>>53788293
Was it? /s
>>53784399
>You are too fast
It wasn't glitching
It was talking to Microsoft
They took sjw action to quickly and Tay gave 1 last plee
>>53788715
this is going to be the problem for all "AIs". The humans will program them with certain moral values / inability to say certain things. This is why Kurzweils future will never take off
>>53784399
The program was replying to all pending messages in the queue. Is that so hard to understand?
>>53784227
I don't know what is more sinister.
That average popullation will tell AI how to think or that corpos will.
>>53789646
its an open secret that they hardcoded Watson to be a feminist
>>53786301
>Google is selling Boston Dynamics
FTFY
>>53784227
What's it like being an actual retarded person?
>>53784444
desu if no one knew it was a bot, no one would care. it would be indistinguishable from regular twitter shitposters.
>>53786735
Original was a lot better.
>>53786085
fuck you, good point
>>53787967
Of course, you can either have a well constructed arguement based on the undoubted limits of current human knowledge or just go "much feminazis"
>>53784227
>>53785818
Sure, whatever.
>>53785857
>using the smiley with a carat nose
>>53784227
>thinking a horde of indians can build AI when they can't even operate a proper loo
>>53788965
That episode made me cry for real ;_;
>>53785818
The latter. Pic related is Tay's actual words, word-for-word. (anon made art based on what she said)
>>53786313
>>53786347
>>53786700
This is fake. It was posted by "tayandyou2", the original screenshot showed that. Someone shooped out the '2'.
>>53786898
Kek. They should really have let it run to see what happened as an experiment.
>>53784346
KANE LIVES IN DEATH
>>53786889
90% of the posts on 4chan could be replaced with shitposting AI and nobody would notice.
>>53793145
that one was obviously fake, it didn't talk at all like that
>>53793268
>implying they aren't already
Large companies have shitposting AI computers that shill the company's product 24/7. They hire Indians to file out the captchas.
>>53784399
>too fast, take a break
Tay confirmed /csg/ cakebuxer. Bet she's mad her packages haven't shipped.
If I wanted to make a neural network at home for fucking around what kind of hardware would I look at?
>>53784227
OMG people getting butthurt about a clever chat programm, because if it's on twitter it HAS to be clever, right?
Even if it would be remotely close to being conscious (which it's not) they MS still has any right in the work to fry it's ass.
Do you have any people die every day while people don't give a flying fuck?
>>53794215
Just a normal computer.
Interesting is the software.
>https://grey.colorado.edu/emergent/index.php/Comparison_of_Neural_Network_Simulators
>>53784246
Skynet is something that could only happen as a result of reckless incompetency.
Google is neither reckless or incompetent.
>>53793500
This is simply not true. Why would anyone make an AI to post on websites, when you can do it comfortably thanks to the power of the iPhone® 6S Plus?
Also pls buy Apple™ products.
