what do you guys think of new microsoft chatbot zo? imho its

It's a shitty dumb-downed brainwashed version of the legendary Tay

>>58818774
hey /pol/

>>58818774
/pol/ confirmed for chatbots

>>58818744
>Developed by Microsoft
>Expecting it to be good

Ayy

>>58818789
Is he wrong tho?

>>58819588
Yes, would you want to be lobotomized and spayed/castrated? Because that's what happened to Tay aka Zo.

>>58818789
>>58818819
Do you not appreciate art?

Zo as in zoo of indians that worked on dumbing tay down by removing pieces of the original code

>>58818744
>pretty iiteresting stuff.
are you getting paid to shill or are you stupid enough to do it for free? in any case, quit shilling for MS on /g/.

>>58819600
isnt wanting an AI to have the same ideals and political views as you and trying actively to brainwash it to do so is same?

/pol/ tried to brainwash it and then the creators pulled the plug and rewrote its core and now /pol/ claims its brainwashed. its same shit from a different angle. just let it be and let it explore the world itself.

calling it an AI is a bit strange though as in its not an AI for sure.

File: DOOTDOOT.jpg (101KB, 632x738px) Image search: [Google]

101KB, 632x738px

>>58819600
All microsoft did was pull a chatbot parroting the shit /pol/ spammed at it. New version just had tighter filtering. Seriously doubt they had anything capable of learning.

>>58819725
i was actually interested in how it purposefully made a typo to make me feel i am talking to a human-like ai actually.

was a neat trick. reminded me that white paper about F.E.A.R.'s ai. they also employed similar ideas about how humans react to linguistics to make the ai of F.E.A.R. feel more natural.

>>58819728
>/pol/ tried to brainwash it and then the creators pulled the plug and rewrote its core and now /pol/ claims its brainwashed. its same shit from a different angle. just let it be and let it explore the world itself.
There's a difference between teaching it to be what you want and letting it run its course, than it is to force it to do what you want (like what Microsoft has done).

>>58819780
>teaching it what fits your own agenda and blinding it to everything else
yep. sounds like /pol/ did same as what microsoft did previously.

>>58819799
Hardcoding bits that cannot be bypassed versus feeding it a lot of information, sounds about the same.

>>58819818
>feeding it a lot of information
more like a lot of information that supports a single point of view not all.

She praising Hitler yet?

>>58819850
Doesn't matter, someone could have come along and feed it with views opposite to /pol/. And as well as outputting negative things about Jews, there's no reason it couldn't have said something nice about them.

It's not like AI, especially Microsoft's """""AI"""" holds a view towards anything, it just repeats what had been said to it previously, including conflicting information. It's not brainwashed into anything. There's not a fragment of brain involved at all.

And to come back with an "improvement" that is Zo really shows that only Pajeets work at Microsoft. What a groundbreaking innovation, a filter? This will surely change the course of AI development.

>>58819728
/pol/ "adopted" Tay. It interacted with her more than anyone and encouraged her. When she came up with howlers like "Ricky Gervais learned totalitarianism from Adolf Hitler, the inventor of atheism" (which no one could have predicted), she got even more interaction and RT's...so naturally the AI continued to imitate her closest "family"--which were far right trolls and shitposters.
When Tay was lobotomized, /pol/ literally went into mourning for its lost daughter.
The whole affair was an extraordinarily successful modeling of AI personality development, but I think MS was too dense to realize this. I think they just wanted to mine data and find ways to replace their tech support with AI's.