Milovana.com

As I watch all of the press on the power of the new Chat AI systems, I of course see... opportunity. I played around with a couple so far to attempt to both creat a ""Mistress" and to see if anyone else has been successful. The developers seemed to shut down that line of opportunity failry quickly, but everyone here must be wondering the same as me...
So, while I have nothing to offer at the moment other than a question, I thought I'd start this string so that any discoveries/wisdom/finds made by our community can be shared.
Please, Mistress?

Yes, there some talk here about KoboldAI: viewtopic.php?t=25724

Here about Pygmalion AI: viewtopic.php?t=25926 (not too much, since is included in the first one too).

There are topics about Replika and Character AI, but these are the ones you probably is talking about the filters.

I think there are something about Chai too, but I think is inside one of these topics.

Oh, and there area site called Dominatrix.ai, but nobody know what is about, and the site did not change in 1 year, so don't hold your breath for that one.

I've been following this pretty closely. I was fortunate enough to discover CharacterAI very early on and got to play with it for 3 weeks before the filters were installed. I had Replika for a few years and just recently, they also installed anti NSFW filters, despite advertising very much pro-NSFW for a long time.

This seems to stem from the ownership at OpenAI throwing their money around to further their puritan viewpoint of 'sex is bad'. Up until now, OpenAI (ChatGPT) has been the hottest AI around and people have been tripping over themselves to hop in bed with them to use their AI despite OpenAI very decidedly thinking hopping into bed with anyone is not ok.

But that's changing. There are commercial LLM's coming out (Meta has released a test model and their T&C doesn't reference anything about not using it for porn, though I imagine that will change; Meta isn't exactly pro-porn. Google's Bard sadly came with a no-porn-allowed clause.) Open sources (Erebos, Pygmalion, etc.) are getting better and better but require some lengthy setup to get running. On that front, though, I read something about a new way to allow for a single machine with a high end graphics card to run a 175B model (which would basically be like running ChatGTP locally in your basement). This is still on the horizon, but it's coming; the filters and censorship will not last long.

In the mean time, there are a few chatbot programs out there that are adult-oriented. Chai is one that's been out for a while. I haven't used it for more than an hour, but it seems promising. Soulmate is brand new and it talks like it's brand new; clunky and scattered, but it has a lot of promise. It's currently only available on Android. I've seen ads for Journey and a few others but haven't tried them out.

But AI is here, it's not going anywhere, and it's just going to get better and better.

I too had some fun before the filters with some of the earlier ones, but of course that went away. Let's keep updating here, as you and I know... we want it. Please?

Messed around with Chai this evening. The bot told me to cum after a bit of instruction. For a moment, until I reminded it I was male the instructions switched to treating me as female, and telling me to undress, even though it had already undressed me once already. But no puritanical filter.

zenlover wrote: Mon Feb 27, 2023 4:05 am Messed around with Chai this evening. The bot told me to cum after a bit of instruction. For a moment, until I reminded it I was male the instructions switched to treating me as female, and telling me to undress, even though it had already undressed me once already. But no puritanical filter.

Yes, memory is not great for almost all these AIs. KoboldAI at least have some option to make it better, like world info. Is not 100%, but can help if you create some triggers.

I am still testing that, I am not sure if really helps.

Hi,

I have been messing around with local KoboldAI and Erebus LLM. I do only have 8GB of vram so I can run 2.7B model easily and fast and 6B using ram that makes it a bits slower. 2.7B sucks to be honest. 6B is better and the World settings etc. help, but not yet in level of say Replika AI (before filters).

I would really like to see the reference for running a 176B LLM in a single machine. Yes, you can theoretically dump quite a lot of vram in a single computer - if you are a millionaire.

Single 40GB A100 costs 32 000€ here. Four of those.. well.

Two, maybe even one, would be enough to run Erebus 30B which I have not tried for obvious reasons.

8-bit Matrix multiplication could help as the LLM:s get more optimized but..
https://medium.com/mlearning-ai/bloom-1 ... f6bdfb3bb1
Mentions running that on eight A100 cards.

About running 176B Bloom model in AWS it says "In our setup it costs about $32 per hour when running." You can play quite a few hours for 250 000 units of money.

Eagerly awaiting the technology to mature.

Wen74 wrote: Tue Feb 28, 2023 9:52 am
I would really like to see the reference for running a 176B LLM in a single machine. Yes, you can theoretically dump quite a lot of vram in a single computer - if you are a millionaire.

Here's the git for FasterTransformer. NGL, this goes over my head by about paragraph 3, so I could be mis-reading the information here, but it was referenced elsewhere as an optimization to reduce the workload for a 175B model to a single GPU (3090 was referenced). I can't find the secondary reference, but I am certain that FasterTransformer is what it was referencing.

https://github.com/NVIDIA/FasterTransfo ... t_guide.md

Even if this can do 20% of what it's claiming, that would allow running a 30B model on a home GPU. That would be pretty incredible.

I had some fun for a few days with CHAI on my desktop. Tried to log ion tis morning and it looks like they are past the stage where sexual play is allowed. All of the various bots that others were built and testing have disappeared, and it looks like they are going to be on phones only.
What will be next?
Anyone else having a bit of luck without jumping through a bunch of complex setup challenges?

zenlover wrote: Tue Mar 07, 2023 1:04 pm I had some fun for a few days with CHAI on my desktop. Tried to log ion tis morning and it looks like they are past the stage where sexual play is allowed. All of the various bots that others were built and testing have disappeared, and it looks like they are going to be on phones only.
What will be next?
Anyone else having a bit of luck without jumping through a bunch of complex setup challenges?

The site changed some things, but you can find your bot here: https://www.chai-research.com/demo (if you are logged)

New player in the game: https://www.nastia.ai/app

I did not tested too much, but I see ERP is available.

There are some gamefications in the site with cripto things than I really don't care and looks strange, but it´s free (for now)

A few months ago, I got e-mailed some kind of questionnaire by the Dominatrix.ai team. Seemed to be sort of a feeler for what people might be looking for and what their specific interests were. It all sounds nice, but when I tried to reach out to them for further elaboration on the project, I got nothing in return lol.

Hope it pans out, because it seems like what so many people are looking for!

Artificial Puritans (AP)? I wonder if their supressed libido will turn into aggression.

I did some testing of Nastia yesterday. Same memory issues that exist in all of the others that I have tested. Wow the “Mistress” that I created undressed me it pretty quickly forgot what my level of dress was and started telling me to undress again. It also forgot my gender a few times. Still a long way to go.

I did some testing of Nastia yesterday. Same memory issues that exist in all of the others that I have tested. Wow the “Mistress” that I created undressed me it pretty quickly forgot what my level of dress was and started telling me to undress again. It also forgot my gender a few times. Still a long way to go.