HTML AI Dungeon 2 - an AI that was fed tons of lewd lines

[Eatyourdessertfirst!]

too much to write here but there are promising developments. from what I've seen and read GPT-3 is shit despite "175 billion" that 175 Billion is all the shit they have collected from Nuance / Dragon Naturally Speaking when they aquired it, when they were previously just backdooring the data or feeding Cortana. As they say garbage in garbage out. I'm probably mixing up OpenAI and GPT-3. So Tibetian rock salt lamp in the salad.

Note to self; acquire a copy of Dragon Naturally Speaking Professional before it gets put into the cloud forever.

I'm not so savy on the tech side but I'm intruiged. What is Dragon Naturally Speaking Professional?

 

[SimulatedFantasyHero]

I'm not so savy on the tech side but I'm intruiged. What is Dragon Naturally Speaking Professional?

Nuance ?_?
before "Alexa..",
before "okay Google" and
before "hey Siri"

You must be registered to see the links

laid out the description of a speech understanding system called DRAGON in 1975.

You must be registered to see the links

In 1982 he and Dr.

You must be registered to see the links

, his wife, founded Dragon Systems to release products centered around their voice recognition prototype.

You must be registered to see the links

He was President of the company and she was CEO.

You must be registered to see the links

was first released for

You must be registered to see the links

, and utilized

You must be registered to see the links

, a probabilistic method for temporal

You must be registered to see the links

. At the time, the hardware was not powerful enough to address the problem of

You must be registered to see the links

, and DragonDictate was unable to determine the boundaries of words during continuous speech input.

Dragon Systems released NaturallySpeaking 1.0 as their first continuous dictation product in 1997.

You must be registered to see the links

very short list.

You must be registered to see the links

even shorter.

You must be registered to see the links

In short, never trust your laptop or tablet ever again. "They" don't have to 'listen' when the AI already has words it listens for.

You must be registered to see the links

you are going to need something a little better then a piece of tape from now on... how about a plug in mic override.

You must be registered to see the links

or something like this.


Honestly, that old headset with the broken mic? probably will do the job perfectly.

 

[Novadrone]

I came here to ask about exactly this. I've been using it as a way to write by just using the Story tool. If I don't like the AI prompt, I hit refresh a few times to get something I can work off of. The start is usually a bit sloppy, but the further I go, the more it learns to play into the narrative, and adds its own ideas in to the mix.

If it's alright, I'm going to post part of what I'm working on now. This is just the raw copy and paste, but I could see myself going back and editing it. It's essentially a story where the main character is captured by cyborgs, and made to serve them. Kind of like the Matrix, but more lewd.

Spoiler: Part 1

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

Spoiler: Part 2

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

Spoiler: Part 3

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

 

[NayNay.]

Thanks a lot for pointing me towards that.
It goes along with quite uncommon input. The more time i spent with it, the easier it becomes to go with the flow and "talk" naturally with it instead of trying to force full control. I had some great time and wish the subscription options and energy refill wouldn't require irl payment data. I wish i could just by a 5$/€ crypt currency payment card to use for things like this.
Anyway, i feel its really fun and the right mix of being able to nudge a story in direction that get filled in by the AI.

 

[Ross_69]

The AI seems to have some problems with context, but it's pretty impressive to see how much it can already do.

Technological progress really is a beautiful thing. Especially when it allows you to make more smut.

 

[JustAnotherHuman7]

This thread has aged very badly

 

[Ross_69]

This thread has aged very badly

Why? It's still capable of producing NSFW content. And is possibly the best free option for that.

 

[Hagatagar]

So, it seems that Latitude wants to get back it's former userbase, by not using the old filter anymore.

You must be registered to see the links

Here some of the more or less important parts (read the original by yourself if you are interested).

First a recap (and kinda admission) of some of the current problems with AID.

We recognize that over the last several months aspects of how we’ve had to approach this problem have frustrated many users. Because there was moderation of unpublished content, users were often worried about what might trigger a flag and whether something they or the AI did could get them suspended or banned. Users reported that they couldn’t feel safe to play and explore with the feeling that someone might be reading their story if flagged.

We also recognize that because the AI can create flagged content on its own, it can be extremely frustrating to users if they are penalized for something that the AI itself created.

Here their "new" approach.

So what does this mean for AI Dungeon? Well, for starters, it means we will not be doing any moderation of unpublished single-player content. This means we won’t have any flags, suspensions, or bans for anything users do in single-player play. We will have technological barriers that will seek to prevent the AI from generating content we aren’t okay with it creating — but there won’t be consequences for users and no humans will review those users’ content if those walls are hit. We’re also encrypting user stories to add additional security (see below for more details).

Essentially, users can do or say what they want in single-player play, but the AI may sometimes decline to generate certain types of content.

Additionally, those barriers will only target a minimal number of content categories that we are concerned about — the current main one being content that promotes or glorifies the sexual exploitation of children.

What kind of technological barriers will you have and how will you enforce them?

Every AI output along with the context will go through a filter that prevents it from generating types of content we aren’t okay with it generating. In most cases since we generate several possible options at once, there will be at least one generation that can be delivered to the user.

The next part is interesting, because it seems that Latitude only wants to go against sexual exploitation of children, and not stuff like incest or bestiality, which OpenAI, their AI provider, don't want their AI to be used for.

What if unpublished content goes against third party policies?

If third party providers that we leverage have different content policies than Latitude’s established technological barriers then if a specific request doesn’t meet those policies that specific request will be routed to a Latitude model instead.

The Latitude model is as far as I know the "classic AI" aka, the dumb one.


tl;dr
Instead of punishing users, the AI will simply stop generating certain content.

Will this help regain the lost trust? Will this even work as intended? I don't know.
But it's definitely a step in the right direction.

Nevertheless, I will keep on not using AID for the time being.

 

[Doorknob22]

Nevertheless, I will keep on not using AID for the time being.

Same here. Whatever I do with a machine in the virtual space is no one's business. I supported them for a few months until they lost their shit (and me as a customer).

 

[JHorcoc6]

Same here. Whatever I do with a machine in the virtual space is no one's business. I supported them for a few months until they lost their shit (and me as a customer).

I agree. I also used to support AI Dungeon, but I cancelled my subscription after my account got suspended because a character used the word "childbirth" in a basically innocuous sentence. Ironically, none of the stories where the mc was a serial rapist in a free use world got flagged, but my character discussing his pregnant girlfriend's upcoming delivery wasn't considered acceptable. Talk about priorities!

I've been using NovelAI ever since AI Dungeon went out its way to lose my business. It still has a ways to go before it is as good as AI Dungeon once was, but it is getting there.

 

[JHorcoc6]

I just created a story on AI Dungeon in griffin mode and tried to play it. It's terrible. The censorship is even worse than when I quit last year. At least they don't auto ban this time though. I started a story that had a free use scenario. My MC then had a back and forth conversation with a woman about the limitations of the free use law, but as soon as he used it to try to actually suggest doing anything sexual to any characters I kept getting an error message that I had "triggered" the filter and needed to take the story in a different direction.

I would write in the female character's age and obviously adult physical characteristics so that the game would know that she was 18 or older and then it would let me continue for a couple more lines, but then it would give the "triggered" error message again. I eventually gave up. It took quite a bit of time and effort in AI Dungeon to accomplish far less than I could in NovelAI without issue.

Interestingly, while my MC was trying to get to the point where he could fuck an obviously adult female character "with huge knockers" and the game kept cockblocking him, the AI also repeatedly tried to generate content about spanking a "98 pound fifth grader." I kept undoing that content and going back to the adult woman but the AI wouldn't let me and clearly wanted me to have a story that would very obviously violate their rules.

For now I'll stick with NovelAI. Maybe in another year I'll check back and see if AI Dungeon is still a disaster or if functionality and reason has been restored.

 

[Ross_69]

I eventually gave up. It took quite a bit of time and effort in AI Dungeon to accomplish far less than I could in NovelAI without issue.

To be fair, you are comparing the free model of AID, to paid model of NovelAI.

It's not exactly a fair comparison.

 

[JHorcoc6]

To be fair, you are comparing the free model of AID, to paid model of NovelAI.

It's not exactly a fair comparison.

The issue is the censorship filter, which NovelAI doesn't have. AI Dungeon (even the free version) is much better than NovelAI in terms of content, but is effectively unplayable now due to the censorship filter. Maybe AI Dungeon would be okay for nonsexual stories, but almost every other sentence that I wrote got blocked if I mentioned anything to do with sex.

When I first started playing AI Dungeon a long time ago I used the free version. I eventually did upgrade to the paid Dragon mode, but for the first few months I used the free Griffin version. I had many very elaborate stories on AI Dungeon and I could generally go in whatever direction I wanted. The AI had a tendency back then to suggest content that didn't fit the direction that I was trying to go, and that was annoying, but I could always change it and still continue the story. Dragon was much more advanced, of course, but Griffin was still usable.

When I started an AI Dungeon playthrough yesterday almost anything I said resulted in the game saying that I had "triggered" their filter and that I needed to take the story in a different direction. It seemed that my characters could talk about sex as a philosophical concept but if my MC actually attempted to come onto anyone the game wouldn't let him do it. It would let me write whatever I wanted, but it wouldn't generate any content of its own to advance the plot, which is the whole point of using it.

When I quit AI Dungeon a couple months after their censorship first started I did because I was annoyed that their filter auto suspended my account for using an innocuous phrase and that I had to generally be careful what I said with numbers and so forth but the actual filter messages didn't really come up very often. When I played AI Dungeon yesterday, the filter message came up after almost every input, so it is much worse than it was before in terms of volume, but this time around it doesn't suspend the account it just says it's not going to advance the plot in that direction.

I have Opus tier, which is the highest version of NovelAI but in fairness it is not really as good as pre-censorship Griffin was on AIDungeon. NovelAI is gradually improving and probably will be as good as pre-censorship Griffin someday but it is nowhere near that point now. For example, AI Dungeon was good at keeping track of different kinds of relationships within the same story. My character could have a wife and kids and have a sexual relationship with his wife and a parental relationship with his kids and unless I specifically wrote anything different it would just assume that kind of relationship. In NovelAI, it is a constant battle to convince the AI that my character does not want to fuck every female character in the story regardless of age or relationship to the MC. He can have a sexual relationship with his wife and the AI will assume that means he also wants a sexual relationship with his daughters, his mother, and any other female characters that he encounters. The only way to avoid that happening seems to be to avoid mentioning sex entirely, which defeats the purpose of the story. The AI also constantly mixes up basic biological concepts and generates stuff about my male MC being able to get pregnant, his daughter being his mother, girls getting pregnant through anal sex and blowjobs, referring to his wife as a virgin even though she has kids and they have been married for years, etc. I very rarely if ever had those kinds of problems in AI Dungeon regardless of version.

 

[Stuffie]

Ever since Laptitude started their censorship, I only had one flag pop up after the AI generated some text. Except for that, I never ran into problems with fetish-focussed conent - even if it was listed in their "no no"-list.

Reading through their current changes, the reason might be that I never played my scenarios publicly? The stories were unviewable for other users. Or maybe my fetishes are just less "crass" than all the stuff you folks are into.

Since I wasn't bothered by being flagged/reviewed, I continued paying for and playing AI Dungeon even after people in here called to abandon ship. And nothing happened to me. *shrugs*

 

[knyght]

I have been playing for a while and the filters are now a lot less restrictive. As an impregnation fetishist I originally found a lot of roadblocks. At one point i had to use the word "offspring" a lot in lieu of the alternatives. But right now i haven't had any filter issues in the last couple of months. I would put a sample but i don't know if it would be appropriate for this thread.
It might also be that i actually add my created characters to the world info with their explicit age and I always set my stuff to private and NSFW.

 

[Yxel]

I played a bit

Spoiler

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

 

[Lyra black]

Is the reason why people got blocked because they didn't have correct settings?
like turn off safe mode

 

[knyght]

I would think so especially as that would start "Contaminating" the SFW part of the AI model.

 

[ririmudev]

I would put a sample but i don't know if it would be appropriate for this thread.

Nope, not appropriate. Nothing but sunshine and rainbows over here. And unicorns, but don't ask too many questions about the unicorns.

 

[TheHighSpire]

This is pretty cool. Just a heads up, it worked somewhat poorly on Firefox for me, but it works better on Chrome.