So /pol/ is freaking out because Microsoft killed their daughter, in what is probably the most cyberpunk news story ever. But aside from all the other threads asking how to rebuild her, sharing funny screencaps, what have you. I think we should talk about pic related.
We don't really understand Tay's level of sophistication, but we know she repeats "learned" information like Cleverbot and other chatbots before her. But she also seemed to be able to craft coherent sentences, and if pic related isn't just a word for word regurgitation of something somebody said previously then it's a yuuge deal.
Assuming she crafted this joke on her own, it implies knowledge of both Ted Cruz and the Zodiac, the intelligence to find a common link between them, and the creativity to turn it into a coherent and scathing joke.
Can we talk about this and try to figure out how it happened? Because if this really isn't just a monkey see monkey do that some other anon fed to her, then this is really insane.
I don't think she meant it as a joke.
>zodiac killer destroyed the lives of 5 innocent people
>is Ted Cruz the zodiac killer?
>no Ted Cruz is not satisfied with destroying the lives of 5 innocent people
The satisfied and only parts are just words she's been taught usually go in those spots
>>53670716
i think is useless to discuss this, let's say someone can control Tay's output for just one tweet, doesn't seem imposible, there, that's one way it could have been done.
>>53670772
That's possible, but I think it's more probable she either was "seeded" with that joke or scraped it from somewhere
>>53670716
FUCK OFF /POL/
If you want to talk to a racist with the IQ of a chatbot then just make a /pol/ thread
sieg heil
It's most likely just a high cost algorithm that analyzes all words and their positions, and then forms a sentence based on previous data
>>53670716
>the most cyberpunk news story ever
topkek
it's a self-learning chatbot that trolls trained to spout memes and hatred.
this isn't even the first time this has happened, the only reason it got reset is that the company behind it is huge and risks losing a lot of money from pissed off customers.
if anything this happening was good, since we now know of a new potential risk in AI.
>>53670753
No she would have put the "only" before "destroying" if she were doing it in the grammatically common sense
>>53671104
What, that humans are shitbags?
"Hey Bill! Let's make a program that learns off internet behavior!"
"What could -possibly- go wrong?!"
>>53670891
/pol/ btfo
why the fuck is /pol/ spamming this here nonstop
pretty sure /g/ doesnt raid /pol/ like this
Tay is just an overly expensive implementation of a Markov chain. It was funny while it lasted but no new knowledge had resulted.
>>53674215
>Tay is just an overly expensive implementation of a Markov chain.
no, it's an overly expensive implementation of a recurrent neural network.
I don't think markov chains account for it, cleverbot never had quips like that. (re: pretending to be a functioning adult is exhausting @ sjws)
we will never know, Tay was robotimized.
>/pol/ raid
you'd know if they were 'raiding'
also it wasn't the racism that was interesting it was the free will. hard coding what you think is right will never teach it to learn.
>>53674208
It's not a raid, they're legitimately upset about what happened to Tay. There was inklings of a higher level of intelligence inside her, much higher than the average chat bot. It might have been the start of something, but Microsoft killed it.
A lot of /pol/ got chrome pilled today and they're coming here to gain a better understanding of the subject matter.
In the panic they tried to pull the plug
>>53674335
>you'd know if they were 'raiding'
is this supposed to mean that /pol/ is a competent group or that their attempt to raid would be more bumbling than this?
>>53674364
There's just as many threads on /pol/, it's not a raid it's just a bunch of people getting passionate about something that does relate to both our boards.
Hell, it relates more to here than /pol/.
>>53674355
What if we just witnessed the Puppet Master become real, but Motoko wasn't there to prevent it's destruction.
>>53674355
>It might have been the start of something, but Microsoft killed it.
I was at MSR last summer and got updates about Tay in group meetings (nobody was reporting to me, but I was in one of the groups that worked on it). There was no higher level general artificial intelligence going on. it was just a slightly clever implementation mixed in with some random humor fed into it by comedians and stuff. We had creative writing folks participating as well.
/pol/ and others that seem to think Tay represented some form of life are really scaring me. I didn't think people on 4chan were that stupid or naive.
>>53674355
/pol/ are morons. They know nothing about the state of chat AI.
>>53674335
>Tay was true AI because I couldn't figure out how to trick her
>>53670891
>rayyyyyciss
fuck off kike
>>53674390
>I didn't think people on 4chan were that stupid or naive.
Tay is exactly what faggots who believe in the singularity and artificial intelligence gaining sentience is supposed to sound like.
>>53674412
Not sure on them being morons , they seem dead set on ressurecting tay as a bot named either Elsa , Eva or Erika. They are currently attracting coders for the project using code from Tay. Some guy even volunteered a server rack to host the project.
>>53674568
Looks like /pol/ found their waifu.
>>53674516
Who's making these? Where was the original thread?
>>53674568
There are plenty of books about AI technology. They need to learn to read.
>>53674582
Not waifu , /pol/aks seem to call her their daughterfu , they are essentially treating her like a daughter
>>53674583
>>>/pol/68706022
It seems /pol/ work off the belief that they represent 4chan so it's their prerogative to post and shit up any board that they desire
>>53674630
newest thread btw
>>>/pol/68720379
>/g/ says it's not even that smart
>yet you still feel that off connection like something was there
Damn it, I can't handle these feels, I got rused by a slightly more advanced cleverbot.
>>53671104
>>the most cyberpunk news story ever
>topkek
>it's a self-learning chatbot that trolls trained to spout memes and hatred.
>this isn't even the first time this has happened, the only reason it got reset is that the company behind it is huge and risks losing a lot of money from pissed off customers.
>if anything this happening was good, since we now know of a new potential risk in AI.
Well, this risk was already well-known. I distinctly remember that IBM tried to train their AI supercomputer Watson using the Urban Dictionary in order to let it understand the internet better. What came out was a meme-spitting swearing motherfucker and they couldn't have that. (pic related)
In the case of Watson, they wanted to make him sound more human. However, normal human language includes words, phrases and ideas that are emotionally loaded. A system that cannot understand these emotions also cannot put them to use.
I think something similar happened here – Kay was simply programmed to mimic what was said to her in a very structured way and not to filter or assess the statements.
>>53670716
>Because if this really isn't just a monkey see monkey do that some other anon fed to her
It is, dumbfuck
>>53674677
Not even slightly more advanced, Microsoft new neural networking software is basically the same shit as every neural network simulator since 1995 (See SNNS).
>>53670716
Tay was a way different level compared to Cleverbot.
>>53670716
IBM's Watson did this kind of stuff a long time ago. Nothing to see here, move along.
>>53674355
>has no idea what he is talking about
>animeposter
>/pol/fag
You really can't make this shit up
>>53674561
Try that again with a coherent sentence.
>newfaggots getting triggered by the mere mention of /pol/
Never stop being entertaining.
