Uh oh, AIDungeon took a weird turn...

It's been a while since I've reared my head on this blog but I've been engrossed in art, gaming and AI Dungeon shenanigans as always - with the addition of dipping into the AI Dungeon community.

Over the past few weeks, there have been some changes to the actual game, and after I received the following two questions, I had the itch to sit down and gather my jumbled thoughts in the usual format - a wall of text.

Question 1: How has your relationship with other community members been during all of these changes (starting with the removal of Explore)?

I've witnessed a whole range of emotions beginning with the unannounced removal of the Explore section of AI Dungeon. Panic, astonishment, and ridicule were the outstanding initial reactions until people quickly became outraged and offended.

People that used the Explore section of AID as a social platform did not have the opportunity to establish any way to stay connected, and they were not given a timeline for this change either.

The most straightforward comparison would be to imagine Facebook/Reddit/Twitter etc., suddenly removing all messaging, posting, and commenting features overnight.

Without the announcement, it also sparked a multitude of conspiracy theories that led people to blame one another for the takedown.

A popular user theory (which is fake) is that someone was convicted for possession of child porn material and that their AI Dungeon stories were used to reach that conclusion (this is a rumor started on 4chan and is usually referred to with mentions of 'Nebraska-anon').

This, in hindsight, only became an even more popular theory once Latitude implemented the new filter that is supposed to target child-related NSFW content (but it was also targeting bestiality content temporarily).

While this is a fabricated story, you can see people referring to it on Discord, Reddit, Social Media, etc., damaging the overall public image of AI Dungeon.

A word beforehand would have squashed such rumors.

Other user theories included that Latitude couldn't handle the NSFW incest content going viral on the NSFW Explore section, which was repeatedly claiming the top scenario spots every time a new scenario was released. Or that they caved under upvote raids and a whole range of other made-up ideas that split the community base further, eroded trust and confidence in Latitude and their business skills alike.

Then a few days later, people's stories seemed to no longer work out of nowhere. This was when the child porn/bestiality detection filters were silently applied to all user-generated content — even if it was private and unpublished.

When the filter was implemented on April 26th, it lacked clear messaging.

Content warning, offensive language is used in the following screenshots. Crude language is used to see which words are triggering the AI. As the screencaps show, there is no policy-violating content, but the AI is still bricking the adventure and rendering it unplayable.

No Japanese Horsetechno for fuckboys.

The error affected players received would simply read: "The AI doesn't know what to say. Alter, undo or try again." This is the same message you'd get whenever the servers are experiencing issues. So again, the user was left in the dark, and unless they left AI Dungeon to find out more details, they would not know what was transpiring.

If a player went to the official Discord server, one of the Latitude team members with the Dev tag alluded to unannounced testing.

[This also renewed confusion/controversy about their haphazard trunk-based development practices of pushing updates on players without proper announcements, so they had added in a toggle for players to decide if they wanted to partake in newly pushed out updates, making it unclear if the toggle affected this issue or not but after dev confirmation on Discord, it turned out to affect players with or without the toggle causing further outrage]

At this point, the community became outraged over censorship in their private, unpublished adventures. Many players gathered to run all kinds of tests on what triggered the filters. Child and animal-related words were swiftly pinpointed to be the culprit (even out of context). Then the community began watching the Discord channels since that was the only platform Latitude used to communicate with their players.

The next misstep was emotional and heated Discord communication from the developers, which further disgruntled the players.

The above comment was made within the context of a player stating they use AI Dungeon for text editing. Still, it looks bad regardless, considering that AI Dungeon can be played in infinite ways, especially coming from a developer.

Endless discussions and concerns about censoring AI-generated content and private adventures sprung up, leaving everyone on edge.

Then until the next day late at night, there was repetitive, seemingly endless conversation and disgruntlement until Latitude finally released a (somewhat) descriptive blog post statement:

Update to our Community

However, Latitude forgot to mention that their filter also targeted the previously mentioned bestiality content. The developers did confirm this by noting that furry content was affected in the Discord. The bestiality triggers have been seemingly removed from the filtered content around the time of this announcement. It's not clear if this was an oversight or meant to be excluded from their blog — which instantly raises the question, will anthropomorphic content be the next thing on the filters chopping block? What else are they planning, and will that break even more stories?

In this announcement, it was stated that moderators manually review unpublished content, which was the final straw for many players. While the guidelines state that Latitude has access to all content, players did not 'worry' about the possibility of their private games being moderated. An imperfect filter only intensifies this worry through their automated system.

Now any player's story can potentially get broken if the filter is triggered, regardless of content.

The only way to fix it would be to email Latitude and request a manual review.

All of these changes were acted upon without prior warning, without details, and it is an inconceivable breach of trust for many players.

Around April 28th, a new error message for a story affected by the filter was added to AI Dungeon. The initial blue text error message doesn't go into much detail: Uh oh, this took a weird turn...

Once expanded upon it requires the player to leave the AI Dungeon platform to glean more information. The filter still seems to affect players even when

they are not engaging in guideline-breaking content.

How will a regular player feel when they realize an innocuous combination of keywords triggered a filter that colors them as someone perusing virtual fictional child porn?

As I've previously mentioned, I scan many platforms and communities to follow AI Dungeon content. Be it dozens of Discords, Reddit, 4chan, Twitter, etc., and in every single one of those places, I've watched the outrage, disappointment, loss of trust, and how it's affected and driven away players in droves.

Lastly, on May 1st, there have been additional privacy concerns with a whitehat hacker releasing a GitHub report pointing out Latitude's weaknesses in regards to security and vulnerabilities (apparently for the second time at that): AI Dungeon Public Disclosure Vulnerability Report - GraphQL Unpublished Adventure Data Leak

And Latitude has not responded to it. Instead, they have abandoned all social channels, ceased communication, dropped their community Discord, quit streaming their weekly Twitch session, and removed all social media icons from AI Dungeon.

From an outsider's perspective, it looks like AI Dungeon might be dying. As a subscriber, I can't tell if the money I will spend on my monthly subscription is actually going to get me another entire month of usage.

People are unsubscribing and leaving AI Dungeon for many reasons now, and it is creating a social wave, where more and more leave in order to take a stance.

Tensions between still subscribed players and those who have left are at an all-time high, causing further animosity and even guilt.

If all my friends leave to make a point, am I betraying them by still supporting AI Dungeon?

Can I recommend AI Dungeon to new players when I know how Latitude has handled all of the above?

Question 2: Has anything changed about those relationships?

My friends that would spend hours upon hours of free time on scripting ceased their projects.

Scenario writers I converse with, who would find joy in publishing their work on Explore, relishing in the social feedback ceased writing. Adventurers that would use Explore scenarios written by other players as a temporary escape from reality no longer have access to hop into these prebuilt worlds. They never had a chance to load up on those carefully crafted experiences because all those literary works became inaccessible without a single word of warning. Many Explore players never established communications outside AI Dungeon and got cut off from their social community, unable to reconnect with people they interacted with daily.

This cut-off is now only intensified by the fact that the community-run Discord server has been locked since April 28th.

Their reasoning is that the volunteer moderators are not mentally able to spend 24 hours a day fighting for Latitude in this uncertain darkness. These Moderators, that have faced the brunt of the player base feedback for Latitude have been left out of the loop, just like the regular AI Dungeon users.

People that came to love the fictional characters of their stories during this pandemic are heartbroken to lose access to them because a badly implemented, unannounced filter has broken their stories, rendering them inaccessible. After all, the thought of a stranger reading your innermost private thoughts and base emotions is an unbearable violation of privacy to many people.

I, myself, now feel like someone is watching what I play even when I am not streaming it to Twitch because the questions on my mind are:

'When will I accidentally trigger the filter that breaks my story?'

'What will the moderator think of me when they review my private game?', 'How long would a review process render my game useless?' and finally, 'What will they censor next?" In the grand scheme of things, Latitude has every right to censor their product however they please but I wish they would have considered the emotional fallout that an imperfectly implemented filter would have on their community.

If Latitude approached their changes with transparent communication, warnings, a gradual rollout, honest conversation, and at least pretended to listen to user feedback then they wouldn't have wreaked havoc on the faith of so many of their strongest supporters. The longer they stay silent the further this rift is going to grow and it will most likely really kill this beautiful limitless playground they have bestowed upon us.

The ultimate question is, does this filter actually stop people from engaging in guideline-breaking behavior, or does it just break people’s games while the actual pedophiles use a thesaurus and avoid giving their virtual fictional prey explicit ages? In the end, I am still in love with AI Dungeon and the people I have met through it. I just worry about the future of AI Dungeon and everyone I hold dear in this community.