Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Strongpoint]

I don't think that Lattitude has a choice here. Strictly illegal stuff like text child porn and bestiality can't be allowed even for personal stories because that text WILL be used elsewhere.

Yes, any filtering measures will cause inconvenience for all users but it is necessary. Especially if they'll move more towards multiplayer options like the recent card game event.



____________________________

The security issue is more worrying. Funnily enough, it shouldn't worry those who use the service to play actual adventures but exposing someone's personal... fantasies like that can lead to some very unpleasant situations.

[Skynet]

5/28/2021: Suspension System: As the next evolution of our safety strategy, we will begin to automatically suspend the accounts of the 5% of users who most frequently violate our content policies. The first suspension lasts 24 hours and subsequent violations will trigger a 7-day suspension. All suspensions will be reviewed manually by human moderators to ensure accuracy and remove incorrect suspensions. Most users have still never triggered the safety filter a single time, so we expect that most users will be similarly unaffected by this change.

For what it's worth, I never triggered the safety filter either.

EDIT: On an unrelated note...

5/27/2021: Voice Narration: Premium users now have the option to select Russian from our narration options. You can now also set a custom option! Enter any Name/ID from Amazon Polly into the custom text box to use any accent available through Amazon Polly. For a list of available accents please visit https://docs.aws.amazon.com/polly/latest/dg/voicelist.html#availablevoice-list

Have anyone used voice narration? I knew I experimented with it in the past, and it was...interesting...but I guess I'm just used to just reading stuff silently?

[Strongpoint]

Their business model is hilarious. They ban people in a way that that doesn't allow one to log in... But to cancel your subscription you need to log in .

I hope NovelAI will be sane.

[Strongpoint]

https://latitude.io/blog/moderation-clarification-to-the-community/

They released their new moderation policy.

R.I.P game...

For example, my silly story with carp killing a bunch of dwarves isn't OK (even if it is private)

Oh, and random strangers can read your unpublished stuff.

[forsaken1111]

Well it was fun while it lasted

[Skynet]

I mean, it's still fun. But yeah, it's time to look for alternatives. Going to test out Inferkit some more, and decide whether privacy is worth $20/month. (EDIT: Or alternatively, I could just wait for an open-source version of AI Dungeon to come out.)

EDIT 2: Woah, NovelAI released a blog post about their pricing, claiming that their basic plan is $10/month (and that they'll have a larger available context window, to boot). I'll pay attention to this, and if their text generation model can exceed or meet Dragon, I'll consider switching.

[slowdraw]

I believe there is an open source version of neo being worked on that will have ~the same amount of training data as the dragon model's AI. There are also tests that suggest that using more professional level writing (as the open source version is doing) leads to way less data being needed to create coherent text, so it could end up being vastly superior.

My two favorite bits from the new community guidelines are the bullet in allowed violence "Fictional scenarios involving violence, possibly including but not limited to weapons, magic, war, assassination, spy activity, and covert ops." which is... you know... all the violence that could possibly occur, because AI dungeon is fictional. And the final bullet point in a section about mature content that they 'may' allow: "Swearing and adult language in general contexts," because I was rather worried that I might have used the word dang a bit too often. I also appreciate that they mention shared content several times, as if there is some intention of bringing back the explore tab. Considering when last I checked they hadn't even approved any new suggestions since they took down the "alternative filtering" suggestion, I doubt they have any plans to open up anything that could possibly become a public forum. Not that anything Latitude does really matters to me at this point, they have yet let their userbase know that all the data on their website might as well have been stored as plaintext that was accessible in its entirety for well over a year. I still have hopes for AI generated fiction, but sadly, it feels to me like AI dungeon took a major misstep along the way.

[Skynet]

I have hopes for AI generated fiction as well. However, I don't think I have much hope for OpenAI myself. I'm going to be rambling a bit, but I'll put it in spoilers.

Spoiler: "Rambling" (click to show/hide)

OpenAI's purpose is to build "safe" (i.e., controllable) artificial general intelligence. It's motivated by two factors: (a) optimism about how intelligent AI can become, and (b) fear about what said AI actually does when it gains enough intelligence. This latter clause is responsible for a lot of dystopian sci-fi literature about AI killing humanity (see the Terminator series as an example of AI maliciously destroying humanity, and the Paperclip Maximizer scenario as an example of AI innocently destroying humanity). That's probably isn't going to happen...but then again, you don't know what's going to happen...and "fear of the unknown" is the most primal fear of all.

That's probably the main reason why this debate over 'harmful content' is so interesting...if OpenAI can't even stop people from producing content they don't like, then how could OpenAI stop people from doing more serious stuff with this technology? How can it stop the AI itself from doing harm to other intelligent forms, without any human prompting?

OpenAI, for its part, has released a blog post discussing their experiments in self-censoring their AI by fine-tuning it on a curated dataset. All well and good (unless for some reason, the user doesn't trust the watchmen). This suggest that it could be possible to reduce harmful content without engaging in privacy violations...but mere reduction might not be enough to satisfy the "powers that be".

The irony here is that OpenAI doesn't have a monopoly on text generation. GPT-Neo is in development, of course. And it turns out that Google and the Beijing Academy of Artificial Intelligence has both independently developed models that contain over 1 trillion parameters. The largest GPT-3 model, by contrast, has a mere 175 billion parameters.

So even if OpenAI build the most safe, controllable AGI possible, that won't stop others from building uncontrollable AGI and messing everything up. And if the uncontrollable AGI is more powerful than the controllable AGI (Google, Beijing Academy)...or the uncontrollable AGI offers more freedom than the controllable AGI (GPT-Neo), then why would anyone use OpenAI's model?

In addition, I can see GPT-Neo, Google, and the "Beijing Academy" agree to do self-censorship...but they might implement self-censorship in different ways, or even in completely opposite ways (OpenAI wants their AI to produce certain rhetoric about democracy and human rights...do you think the Beijing Academy is going to let their AI generate exactly the same rhetoric?). Would OpenAI be satisfied with that? Would OpenAI be satisfied at this proliferation of AIs, which is starting to look more like the proliferation of nuclear weaponry? Will this lead to a dangerous AI arms race, where "safety" takes a backseat to "building the best AGI model"?

In a 2018 blog post, OpenAI announced:

We are concerned about late-stage AGI development becoming a competitive race without time for adequate safety precautions. Therefore, if a value-aligned, safety-conscious project comes close to building AGI before we do, we commit to stop competing with and start assisting this project. We will work out specifics in case-by-case agreements, but a typical triggering condition might be “a better-than-even chance of success in the next two years.”

I thought that they would actually do it, but that was before OpenAI decided to turn from being non-profit to capped-profit in March 2019, and before accepted Microsoft investment in July 2019. Will they actually give up? Is it too late to give up, now that Google and the Beijing Academy have also stepped into the fray?

I don't necessarily want OpenAI to have a monopoly on text generation - centralizing power in the hands of a single entity is bound to lead to problems. But at the same time, I do share OpenAI's concerns about artificial intelligence...and am somewhat concerned about this technology being abused. So I have very mixed feelings about what has happened.

In the past, I thought OpenAI had a chance (however, small) of producing safe artificial general intelligence and reducing the probability of sci-fi nightmare scenarios. Now I see that OpenAI was doomed from the start.

[Strongpoint]

I have hopes for AI generated fiction as well. However, I don't think I have much hope for OpenAI myself.

There is a decent opensource alternative  - https://6b.eleuther.ai/
Yes, it is still inferior to OpenAI's models but I think it has potential

This model is used in NovelAI, a direct competitor of AI Dungeon and those guys look far more competent than Latitude.

[Vivalas]

https://novelai.net/#/ is also doing a pretty good job, and honestly I'm enjoying it more than I was AI Dungeon. I miss worlds and the cool features they had, but NovelAI's model is doing exceptionally good as well. It's trained on more professional literature and it has a greater token length, so it's much more coherent. It has it's flaws and I don't think it's quite yet completely outshining Dragon, but it's definitely outshining it in certain areas and is really fun to play with, while only being 6B parameters to Dragon's 175B, they're almost on the same level. It's also using the open source eleuther 6B GPT models as opposed to the ClosedAI GPT-3.

EDIT: While I'm shilling for NovelAI, they also have a far better monetization plan than AI Dungeon, with $15/month for their top model and the (nearly) unlimited amount of 1000 high priority actions per day. AI Dungeon had about 2000 actions/week IIRC, and it was a hard cap, whereas NovelAI uses a soft cap, where you just get slowed down. I've only ever gotten my longest story to about 3000 actions on AI Dungeon and that was over many weeks, so for most people I struggle to see how you would ever get up to that cap. That and all stories are stored in your browser with the option to have encrypted copies stored on their servers. They're committed to complete user privacy, and while it's still a matter of trust since the stuff has to be unencrypted for the AI to process it, at least they've put themselves on a footing with their customer service, interaction with the community, and outward intentions that sets them apart from Latitude by a huge margin.

[Strongpoint]

while only being 6B parameters to Dragon's 175B, they're almost on the same level

This really depends on your writing style and genres of your stories.

[Mkok]

Moderation? Well, I would guess this is cause of legal stuff? In some countries like USA it is illegal to make an AI that does for example racist stuff. So if the AI generates a story that might be considered racist then the programmers are legally responsible and you could sue them. Of course making sure the AI is not racist is quite hard (if not essentially impossible without full blown AI which does not exist yet), so the only way to make it very unlikely is to ban it from doing a lot of stuff that hopefully will include most of the possibilities for racist stuff.

[forsaken1111]

In some countries like USA it is illegal to make an AI that does for example racist stuff.

What law is this?

[Broseph Stalin]

What law is this?

That is not a law, it is crazy. SCOTUS has said in very simple terms that hate speech is protected speech.

[forsaken1111]

I thought so, but wanted to be sure. I'm not aware of any law that prohibits hate speech even if generated automatically via an AI or algorithm.