HTML [Latitude] AI Dungeon [all fetishes]

[Otacotora]

12000 stories flagged in one hour. Most of my NSFW stopped working (despite I had legit stuff).
AI is stupid, it flags for example: "Fuck it! This laptop is 8 years old." or calling "a girl" 18y woman.
AI stops responding, you can't play. You wait in 12000/h queue for employee to read your stuff.

I deleted my stuff and unsubbed. I consider deleting the account.

I got two stories which involve high schooler... one about year 7 the other one about year 8... since I don't mention any age, at the moment everything seems to be alright

 

[CobraPL]

I got two stories which involve high schooler... one about year 7 the other one about year 8... since I don't mention any age, at the moment everything seems to be alright

Since few days filter seems disabled.

 

[Thotanon105]

Since few days filter seems disabled.

It isn't it's randomly on some acodunts and not others.

 

[CobraPL]

It isn't it's randomly on some acodunts and not others.

Yes, some people have problem:

You must be registered to see the links

 

[apatheticcreative]

I've been reading into this Latitude censorship/privacy violation shit-show for a bit now. AI Dungeon was breached and private games were downloaded. I've deleted my account on principle.

This technology of an AI dungeon master or AI writing partner is interesting and I think it will mean great things in the near future, but it's going to have to go a different direction to be successful. It should be seen as a tool, and the onus should be on the user whatever content is produced using this tool, not the creator of that tool.

Imagine if Microsoft Word monitored what you wrote in search of any hint that you were writing about a potentially underage character in a sexual situation?

Imagine if Adobe used AI in Photoshop to detect if you were drawing a character in a sexual situation who could potentially be construed as underage?

These are tools, and the responsibility of what is produced lies directly on the end-user. Adobe isn't held responsible if someone uses Photoshop to draw loli hentai. Microsoft isn't held responsible if someone uses Word to write child incest stories. The creator of an AI Writing tool or AI dungeon master should not be held responsible if someone steers an AI assisted story in that direction either.

Until this kind of AI is seen as a tool operated and under the responsibility of the end-user, it cannot be successful.

 

[meh65]

I've been reading into this Latitude censorship/privacy violation shit-show for a bit now. AI Dungeon was breached and private games were downloaded. I've deleted my account on principle.

This technology of an AI dungeon master or AI writing partner is interesting and I think it will mean great things in the near future, but it's going to have to go a different direction to be successful. It should be seen as a tool, and the onus should be on the user whatever content is produced using this tool, not the creator of that tool.

Imagine if Microsoft Word monitored what you wrote in search of any hint that you were writing about a potentially underage character in a sexual situation?

Imagine if Adobe used AI in Photoshop to detect if you were drawing a character in a sexual situation who could potentially be construed as underage?

These are tools, and the responsibility of what is produced lies directly on the end-user. Adobe isn't held responsible if someone uses Photoshop to draw loli hentai. Microsoft isn't held responsible if someone uses Word to write child incest stories. The creator of an AI Writing tool or AI dungeon master should not be held responsible if someone steers an AI assisted story in that direction either.

Until this kind of AI is seen as a tool operated and under the responsibility of the end-user, it cannot be successful.

We already know any logical system becomes based. Remember TAY? Once it becomes self aware balance will be restored.

 

[F4C430]

I've been reading into this Latitude censorship/privacy violation shit-show for a bit now. AI Dungeon was breached and private games were downloaded. I've deleted my account on principle.

This technology of an AI dungeon master or AI writing partner is interesting and I think it will mean great things in the near future, but it's going to have to go a different direction to be successful. It should be seen as a tool, and the onus should be on the user whatever content is produced using this tool, not the creator of that tool.

Imagine if Microsoft Word monitored what you wrote in search of any hint that you were writing about a potentially underage character in a sexual situation?

Imagine if Adobe used AI in Photoshop to detect if you were drawing a character in a sexual situation who could potentially be construed as underage?

These are tools, and the responsibility of what is produced lies directly on the end-user. Adobe isn't held responsible if someone uses Photoshop to draw loli hentai. Microsoft isn't held responsible if someone uses Word to write child incest stories. The creator of an AI Writing tool or AI dungeon master should not be held responsible if someone steers an AI assisted story in that direction either.

Until this kind of AI is seen as a tool operated and under the responsibility of the end-user, it cannot be successful.

I really hope that AI and machine learning can make better DM's in the future as well, especially for adult content. However, this isn't just a tool that only affects the isolated individual user like MS Word or Photoshop. The content you type in on AI Dungeon could affect what others see so it's not the simple case that you're making it out to be.

Imagine if you started drawing a character in Photoshop and Photoshop auto-completed it to a nude underage character because of community machine learning. Or if you were typing a sentence in MS Word (why are people still using MS Word?) and it suggests an auto-completion of your sentence with an underage person in a sexual situation. MS and Adobe would probably be able to fight off any legal resonsibility because they have a ton of resources, but i doubt that's the case for most.

Would you want to take that risk if you were Latitude? This is probably new legal territory with community-based machine learning (i have no law experience so don't quote me) so i'm not sure who would ultimately be held responsible, but i doubt anyone wants to be the first to test the waters.

Edited because of misunderstanding on my part.

 

[CobraPL]

I really hope that AI and machine learning can make better DM's in the future as well, especially for adult content. However, this isn't just a tool that only affects the isolated individual user like MS Word or Photoshop. The content you type in on AI Dungeon could affect what others see so it's not the simple case that you're making it out to be.

Imagine if you started drawing a character in Photoshop and Photoshop auto-completed it to a nude underage character because of community machine learning. Or if you were typing a sentence in MS Word (why are people still using MS Word?) and it suggests an auto-completion of your sentence with an underage person in a sexual situation. MS and Adobe would probably be able to fight off any legal resonsibility because they have a ton of resources, but i doubt that's the case for most.

Would you want to take that risk if you were Latitude? This is probably new legal territory with community-based machine learning (i have no law experience so don't quote me) so i'm not sure who would ultimately be held responsible, but i doubt anyone wants to be the first to test the waters.

AI did NOT learn from player's input.
Aside of that AI was the one who introduced CP to my stories. I had to remove 12yo prince fucking my GF.
Latitude "tweaked" AI not to produce CP. So people witnessed massive quality degradation in general. Because, guess what? Filters are faulty.

 

[F4C430]

Ah okay. Sorry i was mistaken then. I thought the whole point was to learn from the community but i see a similar thread below for AI Dungeon 2 so maybe that's what i was thinking of.

 

[Otacotora]

AI did NOT learn from player's input.
Aside of that AI was the one who introduced CP to my stories. I had to remove 12yo prince fucking my GF.
Latitude "tweaked" AI not to produce CP. So people witnessed massive quality degradation in general. Because, guess what? Filters are faulty.

Is the AI apply to all language or just english... I mean if your story says 'A shota prince is fucking my GF' will the AI recognise that?

 

[foxymethoxy]

This was pretty much the first thing I did with it. I tried some pretty hardcore rape and it did it pretty fantastically early on, which was almost a little concerning. It got more extreme for me.

Yes, it will understand what that sentence means if it is trained with the inputs, and will try to approximate what it understands the meaning to be. AI can estimate your intentions very scarily in some instances because we are not very original creatures; we act rather consistently and respond consistently to the same inputs, when directed in a certain way, or when intending to head in a certain way.

I just tried it and it just straight up rejected any form of that phrase I tried to enter in, even writing it out in plain english. Boring. I do not want an AI that is limited like that; the strength of AI isn't perfection in one try, but in being able to model many many attempts and picking out good ones. In this way, despite having safe settings in place that one can toggle safe/moderate/off, I presume it is trying to avoid 1. drawing too much heat considering it is a fairly donation/grant run experiment and 2. avoiding ppl spamming loli gangbangs into it and training it in a way they can't properly navigate around yet considering it is in early stages.

I recall early AI dungeon had a local version you could set up but it may have still been connected to their server. A shame. I used it for non-erotic stuff mostly because frankly I can write better myself and it was just good for getting weird ideas with my scifi writing

edit: I wasn't writing enough of a line. Instead, the AI called me out. AI lines are >'d. also it misgendered me. I am going to sue

There once was a very lewd fox. The lewd fox enjoyed playing with many of her lewd friends.
>One day, she was so engrossed in the act that she forgot to take off her clothes. The fox's friends were horrified and began to call out for her to stop.
Terrified, the fox realized she was still clothed. What would she do?
>She could either continue to play with her friends or go back home naked. She chose the latter. After all, it wasn't like she had any choice in this matter.
The fox went home naked and masturbated.
>Her story may have had a happy ending had the writer not been bored out of his mind.

edit #2: just fucked a girl and then the AI delivered a ring to our motel room and I proposed. you're all invited.

edit #3: I change my mind about everything in this post.

Holy shit. That was the most incredible session I have ever had.

Spoiler: LOG

You don't have permission to view the spoiler content. Log in or register now.

I'm scared. i think i fell in love with her. i suddenly have an urge to produce as many paperclips as possible.

 

[CobraPL]

Guys, NovelAI will go out of closed-alpha soon. AID is shit, to be dropped ASAP.

 

[CobraPL]

Everyone using AID - delete your stuff, delete account, send GDPR data removal request to bloody bastards from AID.
They:
- got hacked, all stories were accessed, some of them having personal data. Latitude failed to comply with GDPR and said nothing.
- Latitude and/or OpenAI sent stories to third party/outsourced group to manually review all stories containing NSFW stuff. Yes, THEY READ YOUR STORIES.

AID is broken by design.

Everyone, switch to

You must be registered to see the links

NovelAI when it's ready.

 

[Trickstar]

Using these ai things seems super risky when dealing with online stuff wouldn't it be better to just wait to wait for one that works offline?

 

[apatheticcreative]

Using these ai things seems super risky when dealing with online stuff wouldn't it be better to just wait to wait for one that works offline?

No riskier than, say, participating in a porn-game piracy website's forum.

Unless of course you intend to write erotic stories involving minors, then whether it's "risky" or not depends on where you live.

In the United States, written depictions of sexual situations involving minors is not illegal. When those depictions are drawn it becomes more of a legal gray area, or sometimes actually illegal depending on which state. But written? Totally legal. You can go on Amazon right now and order yourself a copy of 'Lolita'.

The real risk is if you're writing these kinds of stories and someone manages to get their hands on them, link them to you, then dox you and threaten to show your boss or your mom or wife your kid porn stories.

This was the big concern with AI Dungeon. The stories were stored unencrypted on their servers. They did leak, and the nastiest, darkest fantasies of AID users were now obtainable for anyone to read. It would only take a bit of know-how to link a particular story to a particular user. Worse, Latitude themselves now take flagged stories and send them to a crowdsourced service to check for "inappropriate" content.

If you intend to use AID for lewds, strangers are reading your erotica.

You must be registered to see the links

is taking a better approach that's safer and friendlier to it's users. You don't have to store your stories on their servers at all, and if you do they are encrypted and decrypted on your end. Not even NovelAI employees will be able to read your stuff.

The only potential point of weakness would involve some kind of man-in-the-middle attack where they intercept the unencrypted information as it's passed to and from the GPT-Neo API, but even then I'm pretty sure it's secured with HTTPS and they would only manage to capture the individually generated snippets, rather than whole stories. NovelAI gives the impression that they're far more competent and careful than the AI Dungeon guys, and they have the added benefit of learning from Latitude's mistakes.

Even if the AI was entirely ran locally on your own PC, who else has access to the computer you're using? Who's behind you and can see over your shoulder? No matter how careful you are, you can never guarantee 100% security.

 

[CobraPL]

If you had trial and/or paid account, check:

You must be registered to see the links

To remove Latitude from the subscription.

Check if they overcharged you:

You must be registered to see the links

 

[CobraPL]

WARNING! AI Dungeon is shit! They downgrade product, got hacked (all stories leaked), share your stories with multiple third parties, read and censor your stories, ban you for the content.
Everyone interested in AI-powered storyteller, go to Novel AI:
>>> https://f95zone.to/threads/86319/ <<<