- Nov 3, 2017
- 9,236
- 7,754
What this post says, plus they're run by religious mor(m)ons who literally think they're doing good for the world by censoring private non-con stories involving fictional characters.
- Jun 3, 2017
- 994
- 1,532
There's also
You must be registered to see the links
, which IMO is particularly damning from a purely technical perspective, irrespective of any other concerns.TL;DR: AID was trained on sources containing the kind of content they're now trying to censor with simplistic tools that are unsuited to the task. And further, it's questionable at best that any post hoc efforts could overcome the consequences of this training.
Last edited:
It still seems like the best option if you don't want to pay.
When it comes to privacy, you can just create a seperate e-mail that you only use for AI Dungeon, so it's not really connected to you.
Until a free alternative comes out, it seems like it's wrong to call AID just shit. Nothing is perfect, and in the end majority of people use Google despite privacy concerns even when it, unlike AI Dungeon, does have free, ongoing alternatives. Because in the end paying with some of your privacy for good service instead of your cash is usually preferable.
When it comes to privacy, you can just create a seperate e-mail that you only use for AI Dungeon, so it's not really connected to you.
Until a free alternative comes out, it seems like it's wrong to call AID just shit. Nothing is perfect, and in the end majority of people use Google despite privacy concerns even when it, unlike AI Dungeon, does have free, ongoing alternatives. Because in the end paying with some of your privacy for good service instead of your cash is usually preferable.
- Sep 3, 2016
- 2,020
- 4,050
Sharing your stories and data with multiple 3rd parties.
Banning and insulting for nothing or even by content AI generated itself.
Degrading quality - current dragon is worse than half year ago. Most probably it is not dragon anymore.
"'seperate e-mail " - yeah, and don't pay for quasi-Dragon, because payment leaves traces.
" it's wrong to call AID just shit " - it is shit, absolutely. Worse version is free, but still a shit.
Please go back to topic of this thread - NovelAI
Last edited:
Well yeah, as I said, it's good for people that don't want to pay.
It's not shit when there is literally no better alternative for that aspect.
It is pretty much the topic, since AI Dungeon is even mentioned in the OP, and NovelAI is often presented as better alternative to AID.
But it's still only the alternative for the people that pay, AID is argaubly the best choice for people that don't want to pay, or would like to try out what AI storyteller even is before they commit, since NovelAI disappointingly doesn't plan to have a free version, according to post.
Last edited:
- Jun 3, 2017
- 994
- 1,532
Since NAI requires payment, it stands to reason that anyone who's seriously interested in it is prepared to pay. So whereas you see a choice between free AID and paid NAI, NAI users (and prospective users) presumably see it as a choice between paid AID and paid NAI. And privacy concerns probably factor into the decision. So to summarize:
You: unwilling to pay, not concerned about privacy
NAI users: willing to pay, probably concerned about privacy
Not a recipe for a productive discussion.
Last edited:
Yeah, but not everyone reading this post is someone seriously interested in NovelAI. I speak from experience, I only learned that AI storytellers, including AI Dungeon, can handle NSFW after reading this. From the name, I assumed AID is just for generic RPG scenarios and never got interested.
AI storytellers are still pretty novel and niche, and AID threat is locked, so people coming here may be people who are just interested in the idea after hearing about it first, and knowing that the free alternative is there, and does still work, could be helpful.
EDIT: Okay, nevermind about that, I looked at a wrong threat. AID isn't closed.
Last edited:
- Jun 3, 2017
- 994
- 1,532
- Jun 3, 2017
- 994
- 1,532
A question for the experienced folks: what are your thoughts/experiences on "using NAI (or AID when you used it) as a game" ie. setting up a basic framework using A/N, LB and maybe modules and mostly interacting as a character speaking or taking actions?
---------------------------------------------------
Here's some more resources that may be of interest:
Note: this is a copypasta by someone other than OccultSage, so may not be the 'definitive version'. Trying to figure that out.
This is more to my specific interests:
---------------------------------------------------
Here's some more resources that may be of interest:
You must be registered to see the links
You must be registered to see the links
You must be registered to see the links
Note: this is a copypasta by someone other than OccultSage, so may not be the 'definitive version'. Trying to figure that out.
This is more to my specific interests:
You must be registered to see the links
Last edited:
- Sep 3, 2016
- 2,020
- 4,050
There is V4 engine released, now even better!
- Sep 3, 2016
- 2,020
- 4,050
RUMOR, FYI: EleutherAI is 60% far (~7weeks left) in training 20B model. Target - 150K steps. Pace: ~10k/week.
NAI's work is based on EAI's work. It is needed and NAI fully dependable on EAI work. NAI's work is done after having EAI results.
NAI's work is based on EAI's work. It is needed and NAI fully dependable on EAI work. NAI's work is done after having EAI results.
Last edited:
- Dec 6, 2016
- 87
- 74
- Sep 3, 2016
- 2,020
- 4,050
Cool info!
Models have "parameters" parameter, comparable somewhat to synapses from our brain. More parameters/synapses - the better. Current model is 6Billion parameter, and new one will be 20Billion one.
Service is expected to produce higher quality output - better stories, faking "understanding" something, "having" some plans or concepts, confusing things less etc.
Aside of just bigger model, EAI guys made code better in past six months.
So I am hyped for mire "clever" AI
NAI will be better. NAI is a service using machine-learnin-powered, neurolinguistics, text predictor. It uses "AI" engine made by volunteers called EleutherAI.
Models have "parameters" parameter, comparable somewhat to synapses from our brain. More parameters/synapses - the better. Current model is 6Billion parameter, and new one will be 20Billion one.
Service is expected to produce higher quality output - better stories, faking "understanding" something, "having" some plans or concepts, confusing things less etc.
Aside of just bigger model, EAI guys made code better in past six months.
So I am hyped for mire "clever" AI
- Dec 6, 2016
- 87
- 74
Oh OK... so NAI will update the engine from Sigurd to EAI..Cool info!
NAI will be better. NAI is a service using machine-learnin-powered, neurolinguistics, text predictor. It uses "AI" engine made by volunteers called EleutherAI.
Models have "parameters" parameter, comparable somewhat to synapses from our brain. More parameters/synapses - the better. Current model is 6Billion parameter, and new one will be 20Billion one.
Service is expected to produce higher quality output - better stories, faking "understanding" something, "having" some plans or concepts, confusing things less etc.
Aside of just bigger model, EAI guys made code better in past six months.
So I am hyped for mire "clever" AI
Do you have the exact date when the update will be?
- Sep 3, 2016
- 2,020
- 4,050
Sigurd is also based on EAI work. All NAI models are based on EAI work tbh.
No exact dates, some speculations. It should be around two months.
- Sep 3, 2016
- 2,020
- 4,050
EAI may finish work during Feb2022, NAI will release raw engine as an option ASAP, while preparing finetuned engine will take another 1-2 weeks. (rumors)
- Sep 3, 2016
- 2,020
- 4,050
- Dec 6, 2016
- 87
- 74