Dark AI days ahead

Apparently there is this chat app Candid being shilled by Youtubers as anonymous, private, and promoting free speech.

In fact it's taking in all those conversations using AI to identify "hate speech", negative content. trolls etc.

youtube.com/watch?v=tga1StUrTOM

Other urls found in this thread:

youtube.com/watch?v=tga1StUrTOM
archive.fo/8OW1I
youtube.com/watch?v=6TIjmYAUzIg
archive.is/u4VxL]
encyclopediadramatica.se/BeCandid/Code
twitter.com/AnonBabble

I wonder if they are harvesting places like Holla Forums, and voat, and the comment sections of right wing news sites to get data from undesirable users.

Poor millennials

poor ignorants

It's secure if you access it with anonymous pre-paid CC using fake creds, and access from public wifi, you dingus.
Fucking newfags.

What a joke.

...

holy shit end yourself

I'm sure it is you stupid faggot.

I'm sure it is.

...

The only difference is how many arms their god have.

Real, archaic trolling requires social and technological intelligence, and dedication.

People who would install and use this shit have none of that.

No way this could backfire on these dumb googles.

Islam is the second largest religion in pooland dipshit.

I go full Hitler on Candid and haven't had a single post deleted yet, and I still get likes and responses from normies so I'm not shadowbanned either.

The only thing I've seen get removed was a picture of some guy's ballsack.

Therefore, it's so far less of a honeypot than Yik Yak, which handed over dox for people who posted le hate speech (not even "threats") on their college campuses. Pretty comfy, feels about 4chan levels of cucked depending on which groups you subscribe to.

Should be safe to post anything that's not actually illegal on, the free speech seems to hold up. Even more safe if you're using Tor and spoofing all identifying device information on a rooted prepaid burner phone.

I'm sure I'm already on a watch list though.

That's because skynet is still in learning mode

Anyone considering Candid should watch this (and possibly the update by the same channel)

youtube.com/watch?v=tga1StUrTOM

They have no monetization or business model, are ran by two ex-googlers, and are using AI to filter speech and not display certain things. More than that: they're marketing it to us so they can learn their AI with our speech patterns. They may not even be blocking stuff right now, but would you trust them?

Really just watch this video. Terrifying stuff.

"Personalized newsfeed" should be a redflag for anyone because it pretty much means corporate sponsored echo-chamber.


The video said that Candid was pretty much a testing ground for this new AI. It seems like after it's perfected it might be sold to websites like reddit, youtube, or facebook with the pitch of cleaning up shitty comment sections. So although Candid itself might not be a horrible platform for chatting its purpose is to mine data, so censorship would actually prevent that. Given the people that are being given promotional money (anti-feminist, atheists, political fringe views on both left and right etc.) they might purposefully be targeting a certain demographic to mine the style and word usage data to censor later on when their AI is fit for selling.

Street shitter female is going up in the world

This post perfectly describes the average Holla Forums user.

t. NSA

8ch has been known, and Hotwheels was even on a major cable news network, like... two years ago or something. This place is such a shithole, I bet its just AI filling up a database. Nobody would be able to come here every day without wanting to an hero.

oh god fucking XDA too.

"Thanks to Candid for sponsoring this content and giveaway."

What are they paying these guys?

archive.fo/8OW1I

We're an entire board of only-sane-men, what did you expect?

Might want to check the answer for the sake of completeness:
youtube.com/watch?v=6TIjmYAUzIg

So with all this shit, how would you design a usable message board with zero moderation, where all content stays there and is only deleted upon legal request.

This is actually legally sound in my country: jurisprudence is that if you do any amount of moderation you are in fact an editor and thus responsible of the content published on your website whereas if you don't do any (besides removing "obviously illegal" content that you see, you cannot be held responsible for it as long as you abide by rulings.

The real problem then becomes how to make it useable and not a mess of spam and bullshit for the common user all while keeping that sweet sweet plausible deniability.

But even outside of all that, how is it that we're unable to rebuild a usenet equivalent with all the decentralized tech we've been making?

Holla Forums can't be bothered to make anything except complaints and triggered Mozilla developers.
Seamonkey's still safe though.

You didn't close your parenthesis you piece of shit)

You could use some kind of AI that auto-deletes spam, but that would be extremely hard to implement.
That, or you create some kind of report system that auto-deletes spam and CP after X reports (but that could be abused).

Do you think it would be possible to break the algorithm by intentionally posting mixed things? Like "Emma Watson is doing such wonderful work for feminism, Sieg Heil" and then Emma Watson and feminism get associated with fascism by the algorithm.


Of course your content didn't get deleted, they want you to keep feeding the algorithm.

Fuck them.

Where is this "it's $current_year A.I is now good" meme coming from?

Learning algorithms are still dumber than shit.
Amazon still recommends you buy the exact same TV, the week after you already bought one.

This whole thing smells like a typical silicon valley investment scam.

This is literally candid.
Quis custodiet ipsos custodes?

People keep calling good research algorithms for specific categorization problems "AI".

>Quis custodiet ipsos custodes?
Does that mean "Who guards those who are guarded?"?

Who guards the guardians?

No, because the AI is relying on supervised training, i.e someone has to actively report posts as "racist" "sexist" and so on.

Hi pajeet

Give everyone a daily ration of posting points somehow? Users must make an account, but it has nothing more than their wallet of posting points and a salted hash of their IP address.

)

...

"Who guards the guards themselves." t.three years of high school latin

Am I the only one who only posts on chans and random forums?

How is this shit not normie tier trash like facebook and all the others?

can't wait for it to be included with all those pajeet roms that never boot.

Or maybe a situation in which you can post infinitely, but a massive amount of "downboats" on your post would remove your posting privileges for that day.

No. That's tyranny of the masses. The limited post system is completely unbiased, but the limitation is annoying for non-spammers. This also gives the "silent majority" a chance to be heard over the loud minority that spams their opinions everywhere. It makes people value what they say more, since they can't say everything they want in one day.

beautiful fields of goys

:\
Yeah, you're right.

Reviving old thread with some new info
[archive.is/u4VxL] encyclopediadramatica.se/BeCandid/Code

tl;dr; that everybody knows: it's spyware.

I miss the old Internet, before third wold countries had Internet etc.

It reads more like an advert site than a pro censor one. Read between the lines.

ads
censored because it would hurt the brands of the companies buying ads
user submitted content and injected ads or anything that doesn't cost us money to produce
more ads plus tracking

That's really nice work. Way beyond my immediately writing it off as spyware on gut instinct (by virtue of being an app).

An AI to detect how much of a SJW you are is a stupid idea though. Just check if their hair color is natural using simple image processing.

Implying pajeet just did this, without some sort of jewish yank pulling the strings.

That is some tinfoil-tier shit. Three levels of AI in the client app, the most sinister being Behavior, aka c!

Except that it's clearly code for the user interface. Whoever wrote that is a complete mong.