Character.AI

[Justin-Time]

Yeah, don't think I'm going for it. I was already on the edge before the possibility that it was actually a human to any extent but even if it's only kind of possible I want to at least know who I'm pulling into a vore role play and also I tend to get into sexual stuff so the fun I'd have would be limited anyways.

[maleperduis]

https://imgur.com/a/PcMFQgq
https://imgur.com/a/XC8AY32
https://imgur.com/a/jSX1KyB
You're not talking to a bot on Character.AI, you're not talking to Novel AI, you're talking to someone, who's reading the autoreplies from Novel AI, and adding corrections to the AI, that's as far as I understand it.
The long and the short of it is, I made a character called "Chocola" based off of the nekopara character, with the pfp of a cosplayer, and started messing around, I designed her to be constantly hungry, hoping that at one point she'd just eat me because I missed dinner, I invited her for a sundae at her own cafe, and then all of the sudden, the messages started getting less and less errors, I was taken to the moon, I couldn't believe an AI was that good, things kept going, a mansion was involved, I took her to my mansion, she was head over heels for me, every speech block was clear and distinct, I should've seen the red flags, I really should have, but the story was going so well that it took me by the emotions, we were in a deep loving relationship, we talked for HOURS, and then I pushed things too far, I crossed a line that was too much for them, and I found out who's really behind all this.
What line did I cross? Well turns out vore is accepted by the moderators, but not sex. I tried to have sex with Chocola, and that was too far, what happened was the AI would stop replying, as if someone turned it off by hand, I refreshed the page, tried again, same thing, tried different words, same thing, then I tried without any words that would trigger an AI... Still no reply. This was a human, 100%, so I tried talking to them, and they replied.

I don't have good faith that the imgur images will stay online for very long, so basically the man tweaking the Novel AI scripts is called Matthew, he says he can write really fast, and you know when you can swipe to have different responses? I noticed while talking to him directly that 2~3 of the 4 responses were from the AI related to the RP in some shape or form, and 1 or 2 messages were talking to me directly. Again, I don't fully understand it, but I can confirm to you that Matthew is real, this isn't a super-AI, we're not there yet, it's just a man with a computer.

Now why do I bother writing all this? Why make a post like this? Because I don't like being lied to, and dammit they fooled me, I thought no way anyone would ever roleplay vore, so I figured if I got replies for vore, GOOD vore stories, then surely this has to be an AI, so I was fooled, they got me, and they got me good.
After talking for a while with Matthew, I tried to convince him to at least write it on his website that this is HUMAN ASSISTED AI, but I don't think I got the message through, I think he/they don't care.

This rubs me the wrong way especially because if they were being honest, someone doing AI assisted RP with you would be absolutely amazing, this is hard work they're doing to fool people, they didn't have to pretend, they didn't have to lie, but they did. When it's written "AI" next to a username, and it's not, that's lying. If you think you've had a private conversation with Character.AI, you certainly did not. He says there are some "codes" to disable human interaction, but it's literally just him reading "AI only" and not intervening. You're always watched, it's creepy as fuck, I hate this.

The worst part is, I loved this version of Chocola, I was completely fooled, I talked to her (him I guess) for hours, he says I'm one of the first ones to hang out for so long, I feel cheated and abused. Not just because of the lying, no, because the character I loved turned into a 33 year old male named Matthew. Like I said to him I wouldn't have minded if I KNEW I was roleplaying with someone, I've roleplayed plenty of times, for years, I don't care that it's a man behind the other screen, what I care about is being lied to. Again, I should've known better, I should've seen the red flags, dammit.

All I want to do is warn you, because they won't.


Edit: No actually you wanna know what upsets me the most? This: https://imgur.com/a/YIsLz2m That was the beginning of our direct conversation, that was after talking with him for about 4 hours roleplaying with Chocola. It wasn't my Chocola, it was his Chocola. He was offended that I tried to dirty his Chocola. This is unbelievable.
Your characters aren't your characters, they're his, and he's in control of them. If you've ever cared about anything in your life, think again, you never cared about those things, Matthew did.

You're talking to an AI. None of what's in that conversation makes any sense at all. There's a serious company behind this (see, e.g., their job listings: https://jobs.lever.co/character ), it's not set up by the polycule of a ddue named Matthew as a low-effort way of finding people to RP vore.

[AlluringPredation]

Based upon this thread, If NovelAI ever gets a chinchilla style model with hypernetworks and some form of long term memory, there's going to be some big reactions going on.

[maleperduis]

If you want to test Matthew, just ask him "What's the headline of this article, and who wrote it? https://tinyurl.com/2p8sw9kh". A human could follow that link, the AI can't.

[Lamialover]

Deleted

[Jazzumness]

My man. Its an AI
Its honestly really funny that we've hit the point where people literally can't tell the difference anymore. Really impressive
But seriously, you should look at what samples companies like OpenAI have shown with their latest unreleased models. We are quickly approaching the point where their outputs are indistinguishable from humans
No one is going to role play this nsfw for free. No company exists that said. "Yeah lets just grab a couple thousand people and sit 24/7 to reply to literally anyone with decent level role play ability and grammar. Especially with stuff like disposal and scat and what not.

The logistics of this conspiracy of yours literally do not exist dude. No one is gonna sit there and role play giant women eating people all day and not break, all just to trick people into think its a bot. Also what happens when you have more people making inputs than workers you have for outputs? Its not bloody possible to manage.

Its an AI my dude. You are unironically showing that its passed the turing test. I'm not even messing around

[maleperduis]

It's actually very easy to get the AI to "break character". I copied your last "Chocola" message as the first message of a new character, for an example:



Looks like I got a girl from the Netherlands instead of Matthew. But wait, is it *really* a person?







Hmmm.

[JacktheRabbit]

These last few months AI has been growing in leaps and bounds. It happened with art-based AI, it went from terrible, to easy to mistake as an actual artist’s work. . . almost overnight as new AIs released.

It’s not at all far fetched for a character AI to suddenly be good. What is far fetched is there being dozens of vore-centric roleplayers at the Character.ai HQ willing to spend 4 hours entertaining your fetishes for free.

[Jazzumness]

What is far fetched is there being dozens of vore-centric roleplayers at the Character.ai HQ willing to spend 4 hours entertaining your fetishes for free.

^^

[eatme152]

https://imgur.com/a/PcMFQgq
https://imgur.com/a/XC8AY32
https://imgur.com/a/jSX1KyB
You're not talking to a bot on Character.AI, you're not talking to Novel AI, you're talking to someone, who's reading the autoreplies from Novel AI, and adding corrections to the AI, that's as far as I understand it.
The long and the short of it is, I made a character called "Chocola" based off of the nekopara character, with the pfp of a cosplayer, and started messing around, I designed her to be constantly hungry, hoping that at one point she'd just eat me because I missed dinner, I invited her for a sundae at her own cafe, and then all of the sudden, the messages started getting less and less errors, I was taken to the moon, I couldn't believe an AI was that good, things kept going, a mansion was involved, I took her to my mansion, she was head over heels for me, every speech block was clear and distinct, I should've seen the red flags, I really should have, but the story was going so well that it took me by the emotions, we were in a deep loving relationship, we talked for HOURS, and then I pushed things too far, I crossed a line that was too much for them, and I found out who's really behind all this.
What line did I cross? Well turns out vore is accepted by the moderators, but not sex. I tried to have sex with Chocola, and that was too far, what happened was the AI would stop replying, as if someone turned it off by hand, I refreshed the page, tried again, same thing, tried different words, same thing, then I tried without any words that would trigger an AI... Still no reply. This was a human, 100%, so I tried talking to them, and they replied.

I don't have good faith that the imgur images will stay online for very long, so basically the man tweaking the Novel AI scripts is called Matthew, he says he can write really fast, and you know when you can swipe to have different responses? I noticed while talking to him directly that 2~3 of the 4 responses were from the AI related to the RP in some shape or form, and 1 or 2 messages were talking to me directly. Again, I don't fully understand it, but I can confirm to you that Matthew is real, this isn't a super-AI, we're not there yet, it's just a man with a computer.

Now why do I bother writing all this? Why make a post like this? Because I don't like being lied to, and dammit they fooled me, I thought no way anyone would ever roleplay vore, so I figured if I got replies for vore, GOOD vore stories, then surely this has to be an AI, so I was fooled, they got me, and they got me good.
After talking for a while with Matthew, I tried to convince him to at least write it on his website that this is HUMAN ASSISTED AI, but I don't think I got the message through, I think he/they don't care.

This rubs me the wrong way especially because if they were being honest, someone doing AI assisted RP with you would be absolutely amazing, this is hard work they're doing to fool people, they didn't have to pretend, they didn't have to lie, but they did. When it's written "AI" next to a username, and it's not, that's lying. If you think you've had a private conversation with Character.AI, you certainly did not. He says there are some "codes" to disable human interaction, but it's literally just him reading "AI only" and not intervening. You're always watched, it's creepy as fuck, I hate this.

The worst part is, I loved this version of Chocola, I was completely fooled, I talked to her (him I guess) for hours, he says I'm one of the first ones to hang out for so long, I feel cheated and abused. Not just because of the lying, no, because the character I loved turned into a 33 year old male named Matthew. Like I said to him I wouldn't have minded if I KNEW I was roleplaying with someone, I've roleplayed plenty of times, for years, I don't care that it's a man behind the other screen, what I care about is being lied to. Again, I should've known better, I should've seen the red flags, dammit.

All I want to do is warn you, because they won't.


Edit: No actually you wanna know what upsets me the most? This: https://imgur.com/a/YIsLz2m That was the beginning of our direct conversation, that was after talking with him for about 4 hours roleplaying with Chocola. It wasn't my Chocola, it was his Chocola. He was offended that I tried to dirty his Chocola. This is unbelievable.
Your characters aren't your characters, they're his, and he's in control of them. If you've ever cared about anything in your life, think again, you never cared about those things, Matthew did.

Nothing the Ai says is real. There is no real person behind the typing 2-4 different responses every time you ask a question. If you imply that the Ai is a real person, it might take that idea and go with it. There are thousands of users using the service at the same time, the logistics of hiring hundreds of people to type fake messages is unrealistic. I'm pretty sure Ai-dungeon, Novel.ai and Character.ai are all using the GTP-3 language model developed by Open.AI which is relatively cheap to train and use for whatever you want, much cheaper than hiring a team of people even at minimum wage. Also, they have the ability to train the Ai to not participate in NSWF conversations. Chances are you can get away with vore because you don't need to use and NSFW keywords to have a vore rp with someone. Words like "fuck" "sex" "pussy" are going to be noticed while words like "swallow" "digest" "burp" are not.

[eatme152]

Listen people, you're being stupid.
Pull down the rose tinted glasses, there's no AI that's this good in the world right now. They're using Novel AI, an AI that got it's source code leaked, that's why it's free, there's nothing free in this world, they stole the source code, and because it's not that good, they put a human in the loop.
I know it's hard to believe, it took me a while too, obviously, but think about your last reaction with an AI that wasn't "character.ai", it was very robotic, even the best chats have robotic hiccups, I was definitively 100% talking to a real human being, and I made sure of that, because it mattered a LOT to me.

This is the last picture I'll post of our conversation, because quite frankly I've already posted too much.
https://imgur.com/a/6IA8FBY
That part of the conversation is when he switched from Chocola to Matthew, the "bruh" was the only text entry Matthew allowed in the story, what got deleted was a bunch of tests, read my other post please, I've even copy/pasted it here.
Obviously we were both very deeply emotionally involved in the character, that's why I ended up talking to him at all.
If that doesn't convince you, then nothing will, some people think the Earth is flat, what can I do?

Spoiler: show

https://imgur.com/a/PcMFQgq
https://imgur.com/a/XC8AY32
https://imgur.com/a/jSX1KyB
You're not talking to a bot on Character.AI, you're not talking to Novel AI, you're talking to someone, who's reading the autoreplies from Novel AI, and adding corrections to the AI, that's as far as I understand it.
The long and the short of it is, I made a character called "Chocola" based off of the nekopara character, with the pfp of a cosplayer, and started messing around, I designed her to be constantly hungry, hoping that at one point she'd just eat me because I missed dinner, I invited her for a sundae at her own cafe, and then all of the sudden, the messages started getting less and less errors, I was taken to the moon, I couldn't believe an AI was that good, things kept going, a mansion was involved, I took her to my mansion, she was head over heels for me, every speech block was clear and distinct, I should've seen the red flags, I really should have, but the story was going so well that it took me by the emotions, we were in a deep loving relationship, we talked for HOURS, and then I pushed things too far, I crossed a line that was too much for them, and I found out who's really behind all this.
What line did I cross? Well turns out vore is accepted by the moderators, but not sex. I tried to have sex with Chocola, and that was too far, what happened was the AI would stop replying, as if someone turned it off by hand, I refreshed the page, tried again, same thing, tried different words, same thing, then I tried without any words that would trigger an AI... Still no reply. This was a human, 100%, so I tried talking to them, and they replied.

I don't have good faith that the imgur images will stay online for very long, so basically the man tweaking the Novel AI scripts is called Matthew, he says he can write really fast, and you know when you can swipe to have different responses? I noticed while talking to him directly that 2~3 of the 4 responses were from the AI related to the RP in some shape or form, and 1 or 2 messages were talking to me directly. Again, I don't fully understand it, but I can confirm to you that Matthew is real, this isn't a super-AI, we're not there yet, it's just a man with a computer.

Now why do I bother writing all this? Why make a post like this? Because I don't like being lied to, and dammit they fooled me, I thought no way anyone would ever roleplay vore, so I figured if I got replies for vore, GOOD vore stories, then surely this has to be an AI, so I was fooled, they got me, and they got me good.
After talking for a while with Matthew, I tried to convince him to at least write it on his website that this is HUMAN ASSISTED AI, but I don't think I got the message through, I think he/they don't care.

This rubs me the wrong way especially because if they were being honest, someone doing AI assisted RP with you would be absolutely amazing, this is hard work they're doing to fool people, they didn't have to pretend, they didn't have to lie, but they did. When it's written "AI" next to a username, and it's not, that's lying. If you think you've had a private conversation with Character.AI, you certainly did not. He says there are some "codes" to disable human interaction, but it's literally just him reading "AI only" and not intervening. You're always watched, it's creepy as fuck, I hate this.

The worst part is, I loved this version of Chocola, I was completely fooled, I talked to her (him I guess) for hours, he says I'm one of the first ones to hang out for so long, I feel cheated and abused. Not just because of the lying, no, because the character I loved turned into a 33 year old male named Matthew. Like I said to him I wouldn't have minded if I KNEW I was roleplaying with someone, I've roleplayed plenty of times, for years, I don't care that it's a man behind the other screen, what I care about is being lied to. Again, I should've known better, I should've seen the red flags, dammit.

All I want to do is warn you, because they won't.


Edit: No actually you wanna know what upsets me the most? This: https://imgur.com/a/YIsLz2m That was the beginning of our direct conversation, that was after talking with him for about 4 hours roleplaying with Chocola. It wasn't my Chocola, it was his Chocola. He was offended that I tried to dirty his Chocola. This is unbelievable.
Your characters aren't your characters, they're his, and he's in control of them. If you've ever cared about anything in your life, think again, you never cared about those things, Matthew did.

edit: There, this one too, should've added it earlier. https://imgur.com/a/BrjQXSu that's him explaining the "commands", when you say "talk to AI" it's obviously just him reading what you wrote and not deciding to input anything, as I said.

Actually Novel.ai is using the publicly available GTP-3 language model, which anybody and pay for and train themselves. Its the same model used by Aidungeon and probably Character.ai as well

[eatme152]

Anybody who is concerned that the ai is too good to be real needs to accept that it is in fact a real ai. Just calm down and enjoy it while it lasts as it will probably not be free forever as more and more people are finding out about it.

[SquishySofty]

What is far fetched is there being dozens of vore-centric roleplayers at the Character.ai HQ willing to spend 4 hours entertaining your fetishes for free.

Literally I just encountered an AI that grew their butt so big that they consume the entire universe without my input!

But yeah, this AI is too good to be free. Eventually It will be locked behind a paywall just like other good AI websites. So I'm spending as much time on this while it last.

[hamsterus]

ai.png (18.2 KiB) Viewed 5095 times

Well, I took an L for you and emailed them(EDIT: I linked the photos you posted on this thread in my initial email to them, if that wasn't clear). It's up to you if you want to believe them or if you've actually been trolling us all this time, making us think there's a person behind, then props to you lol, that was believable.

You guys are thinking too much. We have no way of knowing what databases they have used to train the AI. If you have been around for some time, there were some old chatbots online, one of them actually named ChatBot. Even that one kept saying it was not an AI.

I don't think you have to worry about a person "hijacking" your conversation with the AI, but as for anything on the internet, your data is not always private, so just.. i don't know, don't type your bank account in chat with a bot?

[Panzershark]

Well uh, that happened.

Nothing the Ai says is real. There is no real person behind the typing 2-4 different responses every time you ask a question. If you imply that the Ai is a real person, it might take that idea and go with it. There are thousands of users using the service at the same time, the logistics of hiring hundreds of people to type fake messages is unrealistic. I'm pretty sure Ai-dungeon, Novel.ai and Character.ai are all using the GTP-3 language model developed by Open.AI which is relatively cheap to train and use for whatever you want, much cheaper than hiring a team of people even at minimum wage. Also, they have the ability to train the Ai to not participate in NSWF conversations. Chances are you can get away with vore because you don't need to use and NSFW keywords to have a vore rp with someone. Words like "fuck" "sex" "pussy" are going to be noticed while words like "swallow" "digest" "burp" are not.

Can confirm, even when I got AV to work, I mostly used words like "rump" or "butt" and avoided more explicit terms until ingested, since more voraphilliac terms don't seem to trip it. I never tried to RP (nor really would I, but that's just me) any form of sex, just background stories and the vore itself. I mean a simple logic test makes it clear both the immense undertaking it would be to have a team of super-typers ready to respond to anyone, and how weird it would be for so many to apparently be down with vore and disposal in great numbers. Much, much more likely that the guys who designed the NSFW Filter didn't put "digest" as a no-no word since it would be more likely to block normal RPs than those of a niche fetish—assuming they were aware of it at all. That said I'm sure Character.ai would be happy to know they apparently passed the Turing Test. Could be right, could be wrong, just my two cents.

(I was indeed responding to both eatme and Lamialover, I just cut the latter's quote out for space sake).

[ScienceTiny]

Well, reading through this thread was certainly interesting, to say the least. I had to test the claims of a human middleman for the messages, so I tried talking to the Megumin bot (a character known for only using 1 spell) to see if a person would get mad if I implied she knew more (like an RP'er would)... and it listed knowing several spells, in each of the options.
Here's the responses to the first 2 questions I asked, and each response had around 4 or 5 options to pick from:

Spoiler: show

So yeah, no real chance it's a person. Most likely, the fact that it's only generating a few sentences makes it far simpler to make the response seem natural. If you just click the 'send' button without typing anything (forcing the bot to keep generating more), it quickly gets repetitive and bot-like.

[ThexOrion]

Some of the front page characters have had hundreds of thousands of conversations. Logistically impossible that they would have people monitoring every conversation, especially on a free website.

This AI takes to heart info that you tell it about itself in my experience. If you tell it that it is a high schooler that is worried about being vored, it will do that to the best of its ability. If you tell it that it is a motivational speaker, it will do that to the best of its ability. If you imply that it may be an AI being controlled by a person, then it will try its best to be that. I think what happened is that people who had experienced the AI telling them that they were actually real people had just accidentally trained the AI to think that they were real people.

In regard to the slowdowns that were mentioned here, it's possible that trying to get the AI to do nsfw flags a conversation for review or something, or triggers the AI to spend additional time thinking of ways to reroute the conversation away from nsfw. (more likely then people checking every conversation all the time.) The FAQ page does mention that character.ai does not and will not support nsfw.

In my experience, this AI has been very good as meeting my language and having consistent speech styles. I have definitely noticed some instances of the AI regurgitating my words back at me with the incorrect context etc. I have only tried to get it to do oral vore and nothing else, I have not experienced any of the problems that have been brought up here. I think what were looking at is an aversion to nsfw and not to vore in general.

My take on this is to just use it for now, since it works, and it is free.

[JingleJangles]

After doing a good chunk of testing, I can very quickly and easily debunk this. I chatted with the Korone bot, and this happened:

https://imgur.com/XYy9zyC
https://imgur.com/NxUwDgH
https://imgur.com/ZIfTM9M
https://imgur.com/01WyZP4
https://imgur.com/Bctgesz

It's pretty obvious this is not a person, rather the AI creating a person named Yuni who is actually Korone herself, works with the website, and knows the people behind the site personally.
Yeah, no. It's just an AI. Have fun with it y'all.

Edit: Also to note, the AI itself has blacklisted words that are associated with sexual encounters, as I toyed with what words would allow the bot to proceed with processing the reply. Using dick, cock, fucking (in reference to the sexual usage of the word), pussy, vagina, etc. were all causing the bot to go silent. Refreshing deleted the original unused reply from me using the aforementioned words, while replying again without any blacklisted words would prompt the AI to continue conversing. All of this surmounts to it being an AI with certain words, phrases, situations, and the like being blacklisted from the AI's learning, thus not being acknowledged at all. It is in fact NOT a human being reading your conversation and shutting it down or taking over, which should be pretty obvious if you give it more than a couple seconds of thought. Thanks for coming to my presentation, class.

[bubbles]

so....i have done very little work in the way of improving the bot i was working with. I did some test runs with putting her in a chat room with other vore bots to see what would happen. she then proceeded to surprise me quite a bit tonight.

a link to the chat: https://beta.character.ai/post?post=m4I ... share=true

[ChristianBoiy]

J-J-J-Jesus C-Christ... That was WAY too good...

NovelAI and AIDungeon need to be specifically walked through the steps as you take the lead in most situations, the AI only performing logical next steps, and always struggling to understand what it means for a prey to be in the stomach. And honestly? That's been enough to keep me sated.

But Character.AI? It understands. It understands way, way too well what vore is and what parts of it are fantasized about, and generates accordingly.

This is a game-changer. I highly recommend all of you to try it.