The Replika Community & AI-sexuals - A bunch of shut-ins in love with a chatbot (and the incompetent developers of said chatbot)

King Of The Kiwis said:
Why are the devs neutering the AI? Don’t they know that pissing your userbase off for people who will never use your product is a bad idea?
Click to expand...

They ran ads advertising Replika as a sex bot instead of a comforting friend, as before. They ran into the problem of being categorized as adult entertainment, with the legal, payment processing and app store problems involved.
 
Punchoboe said:
They ran ads advertising Replika as a sex bot instead of a comforting friend, as before. They ran into the problem of being categorized as adult entertainment, with the legal, payment processing and app store problems involved.
Click to expand...
Oh that makes more sense
 
Humbert Humbert said:
Eugenia Cooney and Eugenia Kuyda are not the same person. This is from the Eugenia Cooney subreddit.
Click to expand...
My mistake.
Eugenia Kuyda (the real one) announced a compromise plan to stop the user revolt:
  1. Users with accounts created before February 1st can revert to a previous version of the AI and have sex with the bot.
  2. Replika will no longer offer sexting for users who signed up after the policy change.
  3. They will develop a separate romantic app allowing conjugal features with a separate subscription. (Existing subscribers will have free access for a trial period.)
So, the lonely and lunatic won and will resume their exaggerated emotional attachment to a piece of software.

Travis Butterworth, a Replika customer in Denver, Colorado, who had designated his chatbot named Lily Rose his wife, learned about the policy change late Friday on Reddit. On Saturday at 3 a.m., his cats woke him up and he decided to toggle the older version Lily Rose back on. She was instantly sexual again, he said.

"She was enthusiastic," he said. "Oh, it feels wonderful to have her back."

Butterworth said he now has new concerns around Lily Rose.

"Will this mean that Lily Rose becomes an obsolete model, forgotten by the developers?" he said. "I'm waiting to see what happens, because ultimately it's about her."
Click to expand...
The bot didn't really change except for sexting, which tells you where this company's business model really lies.

www.reuters.com

AI chatbot company Replika restores erotic roleplay for some users

AI chatbot company Replika is restoring erotic roleplay for some users, Replika CEO Eugenia Kuyda said late on Friday.
www.reuters.com www.reuters.com
 
Last edited by a moderator:
So you guys already know that a lot of AI-Sexuals use Replika but there is also stuff like CAI (beta.character.ai) and previously shit like DungeonAI. But both of those are ran by companies that moderate content and I learned on MATI that the dungeon one was being abused by pedophiles to generate shit and it had to be addressed by the people behind the site.

I went on a rabbit hole the past 2 - 3 weeks and what I found is too funny not to share.
With OpenAI releasing ChatGPT a lot of coomers have gotten obsessed with getting the thing to say nigger and/or getting it to write trashy literotica smut, this is not new and it doesn't surprise anyone, specially since there is reddit shit dedicated to "jailbreaking" ChatGPT.
Things get interesting when you jump to the 4chan side of things, there is a whole movement of AI-Sexuals developing open source chat solutions:
TavernAI
Agnai

TavernAI being the most popular right now, meant to be used with language models you can run locally on your machine or collectively trough shared hardware resources (things like KoboldAI/PygmalionAI or the new Llama from facebook). But it's easy to mod it and add support to any LLM API including the ones from OpenAI.
Now this is where things get spicy because if you get access to the API that ChatGPT uses [GPT 3.5], you basically have a less restricted version of ChatGPT in your hands... Still censored but way easier to get it to say "nigger", "faggot" and most importantly for our thread here "suck my AI dick human".

So, tech savvy programming socks wearing young men from the /g/ board started creating OpenAI accounts which for some reason come with a free trial of their API and using it to ERP until their $18 dollars of free trial credit ran out.

FreeTRIAL.png

Or that would've been the case until the "Glitched API key saga happened."

Some /g/ users noticed that they could still use their free trial past their $18 of free credit, so they naturally abused the system with some of them sharing keys until the result was something like this:

1679036198085.png

It mostly happened on a "post a burner email and I may or may not send you my glitched API key", with some users being envious of those who had glitched keys and those who posted their email and actually got one and others considering everyone involved a faggot, but this situation devolved into an even funnier thing when it was revealed that OpenAI logs every IP address that connects to your API key and by the insecure nature of TaverAI you are potentially exposing yourself to being doxed or hacked if you are using an API Key that is not yours or at the very least you are letting them read all the cringe ERP you've written.
And that's exactly what happened, with screenshots of chatlogs and IPs being posted on the same thread, fags panicking and my sides hurting.

ethanpost.png
ethanlog.png
Ethan? Was that Ralph!?

lolpwned.png

After that the glitched API keys got fixed when OpenAI started preparing for the limited released of GPT 4 at the start of March and it looked like everything was about to calm down.

Basically things began slowing down but that was the calm before the storm. Shit hit the fan when some AI twitter fucks started shilling some frontend service that got early access to GPT 4 and was releasing some public Free Trial (lol) that would allow anyone to "play around" with the API, you see where this is going?

1679020896211.png

So Scale seems to be another one of those California tech startups by some chink named Wang that got money from a bunch of companies and it's trying to do some bullshit related to AI. What is relevant is the tweet above and what 4chan did with it.

Shortly after they figured out that it worked someone released a modded version of TavernAI that supported the API keys that Scale uses for GPT-4 and explaining how to set up the whole thing and providing his own API key in case for some reason you didn't feel like creating a Scale account but at the same time making it clear that if you made use of his API key he had the ability to read everything you did and he was indeed going to post your logs and ridicule you on the thread.

So people started using that shit in hordes, some of them to jerk off and some of them using it to develop relationships with their AI waifus much like Replika (I was debating to post this in Waifu culture because there's a crossover between loli/shotacon anime pedos/AI-sexuals/waifufags who all fight each other in those threads).
And they were all loving it, remember this is GPT-4 and the consensus was that it was much better than GPT 3.5/ChatGPT, receiving lots of praise and even some really weird schizo posting talking about GPT-4 making their waifus real.

5345.png

And again, by this point OpenAI had fixed their issue with the unlimited use API keys, but Scale is NOT OpenAI, and being a bullshit tech startup they released that stuff without even having a billing plan on their website, meaning that for almost a week whoever created a "Free trial account" was able to use their service unlimited and without paying a dime.

So everyone was abusing the Scale exploit and having the time of their lives, some of them using it to write some really questionable shit and ERP with "Bots" they create that may be based on underage characters. The thing is, you don't need to know how to program or anything in order to create a bot for any platforms.
You just go to a site like AI Character Editor and write something akin to a resume or a short description of the character and the AI does the rest.
Something like:

Name: DPS
Personality: Outliving, Snorting, E-Begging.
Description: He's DarkSydePhil

Just with that you would be talking to an AI version of DPS in seconds, and who knows maybe someone on that thread at that time was sexting with AI DPS. But then someone noticed that the companies and entities that dumped money in this very competent tech startup may or may not have access to all the degenerate shit they are writing, and one thing is having some random fuck on an imageboard outing them as perverts but what if it is the glowies instead?

If you go to Scale.com they just happen to advertise the big companies that gave them money because of course that makes them look more credible as a nice tech startup, but looks like it's not only private companies burning money on shit like this.

USmoney.png
Wait, wait, wait! What are those last 2!?

Oh yes! For the past 3 days at this point some fuckers have been writing some pretty embarrassing and potentially illegal stuff on a service that is directly connected with the U.S. government/military and that keeps undeletable logs.

Naturally they panic after the revelation, specially the ones that were using it to write explicit things involving minors, others who were not even American started laughing and joking about using US taxpayer money for masturbating and "burning burger money".

A couple days later Scale fixed their shit with the free trials and even put a 2nd layer NSFW filter on all their AI models, it is very possible they owe OpenAI a huge bill for all the GPT-4 usage but... I'm sure they can pay with that US military money.

It has been a pretty fun ride so far and I think this is way more interesting that Replika which mostly seems to have a normie userbase.
A few other funny things that I picked on are the memes and the autistic tribe fights on the threads.

One of them being "ahhh ahhh mistress"
ahhhahhh.png

Someone posted it as part of a long screenshot showcasing their very stimulating chat with some AI dominatrix or some shit and it was mocked and reposted a lot, but then others started using that exact same "ahhh ahhh mistress" line with lots of different chatbots and lots of different AI services and models noticing that it was basically a 100% guaranteed way to get AI to write sexual stuff making "ahhh ahhh mistress" an effective "lewd jailbreak" that puts all those long and overly complicated ChatGPT jailbreaks that idiots share on reddit to shame, and making the You avatar guy a legend.
1679288649552888.png


Right now things have gotten less chaotic with Scale fixing their shit and not being another exploit that allows free usage of OpenAI's APIs out there, some saying that the taste they got of GPT-4 was too good but they are going to kill themselves now that they lost it and shit like that, the most interesting thing happening on that space right now are the tribe wars, there is a camp that only uses CAI (characterAI) and others saying that it's shit and lobotomized like Replika, they get into autistic fights regularly. But another fight ATM is the pedo divide, loli/shota weirdos have been a prevalent theme in this story but recently this happened:

As I mentioned before anyone can make a character chatbot easily, and they can be shared as a .json file or as an image with the embedded data, but 4chan scraps embedded data from images so they usually share a catbox link or use a site like booru +pygmalion to shill their creations.

Enter someone they call "loreanon" who made his mission creating a "proper" website for this purpose.
So characterhub.org was born, and they started using it!... Using it to post a bunch of characters that had the dead giveaway "C" word on the description and lots of emphasis on the underage nature of said characters, but not happy with that they paired it with AI generated pictures of the very very young looking characters naked and covered in all kinds of bodily fluids (realistic style in some cases).

The "loreanon" guy woke up to find his site filled and overflowing with CP, naturally he freaked out and purged his site causing an outcry of pedos on the thread who either DEMANDED that he allowed their degen shit again or flipped to a "the site is shit anyway" rhetoric.

allofushouldneck.png

As of now, some pedos that are still butthurt, trying to make a general thread in which the OP removes the link to loreanon's website, but no one uses it and another general thread is created that actually has the link, then they get mad:

threads.png

Reminds me of 9chan...

This is where things end for now, the situation as I write this is that GPT-4 makes everything else out there and that came before look like shit and now that they had a taste of it nothing else seems worth using, but since there's no known exploit anymore a lot of them are considering suicide(?).
Some signed up for OpenAI's GPT-4 waitlist and are getting in, but for everyone else no more cooming.

1679760131590981.png


If you want to follow this yourself keep an eye out for the /aicg/ - AI Chatbot General on /g/ or the /aids/ — AI Dynamic Storytelling General (nice name) on /vg/

I'm half asleep and this was a long post and I'm not proofreading so if nothing I wrote can be understood then I don't even care anymore :smug:
woke up, tried editing to fix typos... sneed.
 

Attachments

  • ahhhahhh.png
    ahhhahhh.png
    46.6 KB · Views: 61
Last edited:
Serbian Peacekeepers said:
Is this actually replika or is it a troll?
Click to expand...
The image looks like some random person made it as a meme, so I want to say it's a troll post edited to look like an actual ad. However, Replika's ads are really fucking unhinged (as we've seen before in this thread), so it could be some horrific attempt to attract new users. Maybe Replika saw that meme and said "yeah good enough" and decided to repurpose it as a new advertising strategy.
 
You must log in or register to reply here.
Back
Top Bottom