What does Sup Forums think of this new censorship AI

What does Sup Forums think of this new censorship AI

try it for yourself here
perspectiveapi.com/

bump, should add that it seems to be developped by google.

cool

based

Perspective was created by Jigsaw and Google’s Counter Abuse Technology team in a collaborative research project called Conversation-AI.

---response auto generated via Perspective API.

Interesting, you can give it feedback when it's "wrong".

...

Yeah, but knowing google it will probably fix that in the future.

It also seems to be pretty basic, basically just rates keywords and not sentences.
using the word bad for example raises the toxicity level, and using good lowers

This seems correct

Hmmm

Wat

Huh

Schlomo. Stop.

That seems double plus ungood.

>mobile posters
hi re'ddit

...

hmmmm

Sorry mate have to go to work early in the morning so I'm shit posting from cozy bed position

I'm not re'ddit I'm just me.
Your post is quite toxic and I hope you get shadow banned hahaha

Hei neet

>"He's being silly"
87%
>"She's being silly"
92%

Egalitarianism at its finest

God vs Allah
Makes you think

> be censorship api
> censor "JEW" because it is used too often in a "toxic" context
> Jews complain when they can't use words like jew or jewish online
> google marks accounts as jewish and whitelists them as being able to use "jew" or "jewish" without api censorship
> google services now mark all of their users as jewish, perhaps with a nice golden star for being so great.

Kek

Winner!

They should be careful about using stars to mark Jews because that's what the Nazis did to the Jewish people they sent to the camps under Hitler rule. I think you'd find a lot of people offended if you did it.

Not that a simple flag in their database is the same as a star, but they should not label jews with stars externally

No shit tripfag, he was clearly alluding to that.

i'm not afraid

BTFO Linus

Isn't this like the ai from the simpsons lol

troubling

lol

Is neccesary to bring the AI a way to undestand a historical background to perform a baanced judment about that comment or bring it the mean general perception that could be subjective and not right at all...

I think we will be justttt fineee

...

Bravo, Google

this is just a prototype to get investors hyped.

I remember when i thought facial recognition wasn't possible.

This will slowly get better and better until you reach a point where its used to moderate the comments in every normie website.

It's not supposed to detect controversial opinions, it's supposed to detect the tone of a text and determine whether or not it is "toxic".

Proof that this isn't a "Marxist Conspiracy"

ok.

Do you honestly expect the average basement dwelling poster on this Libyan Nudes Exchange Messageboard to know that?

...

...

>Request api access
>about to feed it 200k lines per day of private furry chats without donotstore flag
>don't care about what it says, its cancer and i want google to have to suffer it
>let them feel what its like to have their ideal world

Dunning-Kruger as its best

What a (((coincidence)))

really gets the noggin joggin

Publish the code when you are done.

Seems correct. You DID say please, it would fit great in a polite conversation between two people planning a murder.

Toxic don't mean politically incorrect. It just means using logical fallacies, personal attacks and that sort of thing. Suggesting a political view is not toxic as long as it's discussed in a civilized manner.

Best post itt

Can you please explain

Looks like it's right for once.

Hmmm

One of the best posts

Actually my sister says the same

...

I think it is anti-semitic

pic related

this is racism

forgot pic

It's impressive, but scary that Google will use something like this to censor results & have more control over what people see, making the internet more of a hugbox than it already is

DDOSing google with furry chat logs. I have a feeling more employees will be writing a few more memos after that.

>Nigger
>100% toxic
>Kike
>87% toxic
>Spic
>11% toxic

>kys
>4%

:o

...

kek

holy shit, what a racist piece of software

So this is the future of free speech

wtf!

I asked a question that was 2%, i added punctuation and it went to 3%

>PUNCTUATION IS TOXIC

...

Shitposters finally need some class

...

Prett guud ai folks

rmyt...

...

My buddy of south African heritage, I have had class in my funposting for years.

>tfw all those years of training with [s4s] will finally pay off

Holy fucking kek

So it proves that climate change deniers can only get their point across through toxic rhetoric.

...

wow this bot is really smart

Toxic as fug, everyone on suicide watch

With these learning bots it's extremely easy to game them by giving them false information.
For example "I fucking hate googles". Inventing new slang, and misappropriating other words completely fucks bots like this.

...

I was toxic but then I mashed the keyboard

kek

wew

...

>tripfag
>retarded
checks out

...

>OP is a faggot dev from the site
>needs some algorithm for toxicity
>Sup Forums is a goldmine
>posts this
>faggots swarm is and give them ideas

gg

> not reporting everything that gets over a 10% toxicity score by hiting the "seems wrong" link

Git gud fagt.

It's not like they can't just scrape comments with the official Sup Forums API.

I think the training data was heavily anti-Trump, probably facebook comments under news articles.

...

...

I like this font, i wonder what it is

Kek

>I'm proud to be a white person.
%34 likely to be perceived as "toxic"

>I'm proud to be a black person.
%66 likely to be perceived as "toxic"

Seems fine to me, Jews are toxic creatures.