Post

Had a quick read, seems like a very reasonable stance to be in considering the scale of the platform, but I'm curious how people will react. Glad to see it's being addressed though. Finally something being done about the slop world thumbnails at least.

I'm actually ok with this. For now at least. I do understand the grey areas with AI. But I do strongly feel that AI disclosure tags are still necessary. If I visit a world that has AI-gen parts or code, it should only be because I was informed ahead of time. I don't like being deceived.

Regarding tag, I believe tags should come with additional context. Details such as how much of the scripts were vibe coded, how much of the assets were AI-gen, how much manual post processing was used on the AI-gen parts, and so on.

I'm posting my thoughts here because it looks like the survey requires a Google account and I will not sign in to Google just to participate in a survey.

I will note, as also pointed out in the post, that tagging is a cumbersome ordeal. If we rely soley on the content "creator", a lot wont put tag on, knowing it will hurt their content exposure. Having VrChat somehow police every world, would probably bankrupt them with the amlint of ppl needed.

scorched earth take: what is the point of business development velocity mindsets in regards to worlds development in a platform like VRChat? it's the same perspective I have towards the increasing LLM usage in the game recompilation scene. this should be an opportunity to learn, not pump out content

do we really need new worlds at a faster rate? my friends spend a ton of time in worlds that have been around for years because there was consideration to their creation. why are we even here if not to share expressions that demonstrate artistic/technical skills and the passion of imagination?

I don't share this plucky utilitarian idea that 'some' use of the planet-killing, mass plagiarism engine is okay. there are hundreds of potential translators you can find. there are thousands of skilled coders to help with scripting. connect to those talented people instead of trying to replace them

I'm really glad there is a response and a road map, but I do have some issues with the survey as I'm currently filling it Some categories in there being merged into the same question I feel should be separated (I.E : meshes/finishing touches)

Maybe it's to help more people read the questions more easily? But I find that lumping some of these together makes the percentage of AI generated/assisted content more ambiguous, which in turn is going to make my choice harder for which tag I'll choose.

A mesh could be the whole terrain, audio could be one voice clip or the entire sound design of a world. Heck, most worlds don't even use audio. So usually adding it would go beyond a finishing touch (like voice acting in a game) And yet these are all the same category.

For now, I'll try to follow the standard set by VRChat's own avatar performance rank and go for the most "pessimistic" option for each question, it'd follow with how safety features are handled and it'd be more fair to those who would want to avoid this kind of content altogether (including myself)

Please, oh please, next time tell us that you're preparing a response instead of being silent about it? Being silent about a topic for so long makes it look like you're ignoring the problem and hoping it goes away.

The Google form seems to want us to have the conversation of "How much AI is too much AI", but that's not the issue for me. The issue for me is "what kind of AI is being used". Because some do disproportional harm, and others don't really do any harm Answered anyway, but yeah

Would you be willing to apply the same standard you pick for users to yourselves? I don’t condone your use of AI art, even internally. You host a community of artists. I don’t mind LLM coding, that’s distinct from genAI art, but apply it fairly. Allow categories 1-5 for creators and VRChat equally.

Curious why coding gets a pass? I see this all the time - people want to protect artists, but developers are thrown under the bus. LLMs can code for the same reason that they can create pictures - they've stolen billions of lines of code to learn, with no attribution or compensation.

AI art fully replaces the artist. If you generate a fursona instead of a commission, that’s direct job displacement. Code is not like that. Software is never finished, tools that accelerate development don’t replace programmers, instead they improve their speed and capabilities.

As for the training data argument, largely it is the same story. Artists almost universally do not condone training on their art, while in programming, open source and the free sharing of ideas is commonplace. There is of course complications and nuance, like licensing concerns, but that’s the gist

I'm my experience LLMs are absolutely replacing the developers, at least in enterprises. So it's weird to me that people are cool with vibe-coding. Oh, it's just a tool to speed things up? No, CEOs see bogus stats like "43% efficiency" and they will absolutely fire 43% of the workforce.

I agree with you about stealing code to train models. I'd say that, at its core, code is more of a utility than an expression or a form of communication, so maybe that's the reason for the pass. But that doesn't excuse the theft in the first place.

I feel like this Talks about tagging, and the form just talks about whether they need a tag, but it doesn't go into whether fully AI worlds should even be a thing or whether AI content should be allowed on the marketplace. I also worry about AI bots filling out the form.

Thanks for asking our feedback. The worst part about AI is that because I'm low vision, it's a lot of convenience for things that help with independence. But gen AI is everywhere and I'm bordering on afraid to do anything art-related for fear of surprise AI and awful quality. Now in VR too?!

I hope the pushback I've seen on social media actually translates to meaningful feedback in the form. at least there's now a good way to leave feedback, albeit way too late regardless I've left my thoughts in there because there's no way to fit them in 300 characters lol

My thought has always been VRC should closely align with Steam's AI disclosure policy (both as a good model of practice and covering the company's butt since we are on the platform, though user generated content might forgo the need to label on steam store page?) a good first step to address this ty

You know, this kind of thing will be a lot easier to talk about once all these ai companies stop trying to kill the world with their evil evil buinsess practices. And until that day comes I would simply refuse to play VRChat if any level of ai is involved.

This isn't really related to the post at hand, but I find myself unwilling to register since your site still blocks Protonmail addresses from registration. I don't want to use a google email (or frankly any other popular provider), but I would like to play again. Any idea when this will be resolved?

So now that the Creator Guidelines were updates, how do we report misleading content? What's the proper report option?

Please also consider letting us sorta curate our own world discovery systems, like following specific creators or following someones feed of curated worlds. Adjusting our own experiences, like muting/blocking someone's worlds from showing up. Etc.

If you think the discoverability problem is bad now, do not open the door to the FLOOD. While I don't think any usage is ok, from an actuall enforcability stance i think everything yall are confident about is sensible. But the three items you have debate over I would encorage a no on.

Merely for A) discoverability for actual creators would crater given the flood of generated slop, and B) theres a substancial thread to genuinely having to host that much data with no limit on how fast that shit can be reproduced. Its a hosting threat too.

Here's your solution: Get rid of Generative AI. Yeah that includes worlds that make it "look good" too. I don't care. It's anti human and anti art and taking ANY stance other than "Fuck Gen AI" is ignorant. Make it a tag system. YES. For everything. Even textures and code. For groups & avis too.

This is great, thank you for addressing things and giving us a form to fill out for feedback. That's to also say to everyone in the comments, fill...out...the...form. That's how your voice will be heard! Much love

I'm glad you guys final came back to us on this. But all we ask is just even a" hey we hear you guys give us time to respond and work on something" not to rush it. Iv filled out the form I hope you guys make a some changes from the fed back, thanks again !

I dunno man, willingly using AI just tells me you're too lazy to learn how to do something right. Where is the spite power to learn?

This, this exactly. I've run into so many walls and flailed around aimlessly so much when making my current world. It's been over 2 years since it was just a floor plane with a box, and I'm not even at v.1.0 yet.

This is what we all wanted. A response. Please next time, let the community know you at least see us. The response is well written, and tries to include both sides. Appreciated. Even though I don't like AI. The mass production of AI assets and worlds, just became far too much to ignore.

Let's give everyone here a little grace as we ALL work through these issues. People's opinions can evolve and will likely change as arguments are made :)

Thank you for listening to our concerns. We all know this isn't an easy problem to solve. As long as you're making serious efforts to address the AI slop problem, we'll be right there with you. And hopefully, in the long term, this will help fix discoverability and world feedback issues!

I like what I've read and I like the immediate changes that are planned. The question now is what the follow through will look like once all of the data from the form has been collected. We'll be watching. Please, please don't ruin this place that so many of us call home.

This is a good response, thank you. I hope the team doesn't simply look at the checkbox/Quantitative data, but will make sure to listen to the worded responses/Qualitative data. Too often has the larger community been unheard, only for a select few be prioritised for "the numbers".

Dumb question: Does all of this force @vrchat.com to have disclose AI content on their steam listing? Does all this start to cause problems with Steam itself? Are they currently breaking Valves TOS?

I mean From what it sounds like VRChat should probably have an AI use disclaimer on Steam stating how AI has been used in an assistive way for coding. + also explain how users may upload UGC with AI assistance or AI generation (once they decide what they are allowing users to upload, going forward)

No, there's no way Valve would consider UGC under the content disclosure. If it did, every single game with UGC would need it, which would make the disclosure useless.

This. I'm not a lawyer, but yeah, user-generated content isn't the same as what exists in the platform. That's wider than Steam policy; look at section 230 in the states for example.

Steam can slice that differently as a private entity, but I think it's guided by the same notion that hosted content is not the same as the platform's code or speech.

In anything submitted by players: No As placeholder assets, like the one that made it into production by mistake: I assume so, yeah

Thanks! The silence had me scared. I understand why it took time. Its an important topic. At the end of the day AI exists only because of theft of human efforts. It has no value in my life. Use it for yourself I dont care, but dont show it to me like you ”made” something of value.

the domain is blocked by dns ad blockers, so make sure if you're using one it's disabled for 30s or w/e.

I can't afford a new PC because of Ai, so I cant play your game. Any embracement of the technology is positive feedback for them keep destroying the home PC build market.

I look forward to seeing how much of this is talk and how much actually gets acted on! The silence felt very wrong; I think acknowledging it earlier would've been nice, even if it was just an acknowledgement. Still, I'm happy to see a few of our real, deep gripes stated as fact in that post.

I really want both cookie cutter worlds and fully generated ones to be hidden by default for everyone. Nobody should be sifting through that much garbage unless they explicitly want to.

please do something about the rampant AI usage. it's incredibly disheartening seeing so many VR chat worlds using AI assets everywhere, whether it be paintings on the walls, AI videos played on media players, or AI generated music, some worlds are incredibly rampant with AI assets and no disclosures

This is promising. A response like this is really what most of us were looking for in the first place, and even an "Updates on this soon" would've put most of us at bay for a bit. Keep up this pace. This should be front-and-center of your team's full attention (within reason) for awhile.

What is there to talk about? All generative AI has been trained on the stolen work of others, most of which are small time creators. It's been the largest theft of creative content in history. Generative AI = Theft. Plain and simple.

These companies love moving the goalpost to get away with the exploitation of others without losing credibility. The whole “we’re using AI responsibly!” approach when there’s no such thing.

Glad you all finally responded. I do hope our feedback is taken seriously and the dialogue doesn’t take months or years to see some change. This is a very serious matter and it needs to be handled properly.

Thank you. I hope that the "dialogue must continue" isn't a dialogue with months or years in between, which has often been what I've observed with VRChat's dialogue, even before the AI elephant.

that's why we have the 6 weeks timeline in there, discoverability is hard to measure without someone to look back at you when you start turning the knob

"Some of us use AI to automate parts of our work or fill gaps in our knowledge" alright then I'm not using VRC

Please please fulfill your promise of listening to us. As a creative worker, even worked with furality, more often it feels like our thoughts and feedback is completely ignored. Filled in the form, so thanks for at least letting us submit our thoughts.

Literally just stop using ai period in your work! 1 grammar and translation, this can be fixed with having someone who can translate or idk maybe a book? 2 coding, again if you can’t fix it find someone who can, stop relying on the easy cheap “solution”. Ai isn’t the answer, fix it yourself.

I have bad news: the operating system you're using right now, the app/browser you have open reading this, and the Bluesky platform itself use AI to some degree both in code and functionality, meaning your message is AI assisted. That's the problem. The "line" isn't simple, hence the need to discuss

“And yet you participate in society” response and logic. It is clear cut. Just because the tech industry like to play non-consensual “sneaky link” with my computer and apps doesn’t mean we can’t oppose AI.

And don’t ya think that in it of itself is a bad thing, I truely have never seen ai in a good light whether generative or not.

and yet you participate in society ass answer. The only blurring of the line is that the slop merchants have successfully conflated everything that is not literally hand crafted into a shapeless blob of "AI" instead of distinguishing LLMs from procedural tools and algorithms.

See the context of the message. Part of the problem is that conflation causing people to say things like "just stop using AI period" to the topic of having a discussion of "the line". What kinds of models vs algorithms? What does it mean to "use"? Those are relevant questions to the discussion.

I whole heartedly agree, companies need to stop relying on ai, especially when it comes to fixing their human shortcomings. Just hire someone worth hiring and pay fair wages for their work. It’s all that we have been asking for!

If a message is presented exactly as it was typed, then AI didn't assist them in writing their message. "Writing was assisted by AI" Typically is understood to refer to AI umbrella tools being used to alter or generate text.

One may not be happy knowing the platform was built using AI, for a variety of ethical and practical reasons, but the meaning of "This message was AI assisted" is not genuinely ambiguous to most people who are discussing this.

The line is very simple. The awkward thing is that ever since BlueSky team gloated about using AI, the site's been less stable. Nvidia's drivers are terrible, Windows has so many issues lately since they went AI coding... seems like there's reasons to think it's bad.

Most creators don’t have budgets. While on principle, I agree, it would be preferable to pay professionals, I don’t think it’s reasonable to expect most VRChat creators to be able to do so. Studying translation by book and fixing code by hand is work on the scale of multiple college degrees.

So for 1 & 2, for many creators, this isn’t a question of “AI” vs “handwritten”, It’s a question of whether the translation/feature exists at all. If there’s no job displacement, and the code/translation isn’t a thing you ever could have done by hand anyway, I don’t think there are real harms here.

A suggestion I give people is make what you can, put a sign in your world asking for translation help or whatever else you need. Kind and creative people are out there and will let you know if they want to volunteer their time for something they think is worth it.

While you do make great points and on the topic of budget, I still agree, my comment was more targeted towards vrchat and corporations vs the small creator. However using ai in your work imo is still not an answer or solution to get the work done.

What makes you dumbarses think putting ai GARBAGE into vrchat was a good idea? The one that thought of that should be fired right tf now, alongside everyone that supported this utterly stupid idea.

I'm glad we finally have at least some word, but dragging your feet on a tagging/detection system now because you're afraid of having to upkeep said system in the future solves nothing.

Tupper being subscribed to a AI Discord and website that is specifically only for generative AI images and stories and nothing to do with coding....

That, along with this paragraph, suggests to me that VRChat might've just made up their minds. "High quality" AI-gen worlds are just gonna be a thing, and VRChat really doesn't want to hinder their success, so those of us who're opposed just have to live with that. I'm shocked people don't see it.

No matter what they need to allow us to have a AI label system. That is a hard need for me. If you are worried about people dogpiling them for using AI then you need to understand that a majority of the community does not want it and deal with it.

The devs need to understand that no matter their own view on AI. If the user base doesn't want it, then it needs to go. Or give us the tools to filter it and be done with it. If people are made fun of or dogpile because of it... Sucks to be them.

A cleverly staged thumbnail might be visually accurate to the world and still hide the extent of AI generation. It may fail to raise a flag for misrepresentation, yet still lead a player to an undesirable experience.

Ranking quality, as with every discussion about "prioritizing quality art and banishing low quality art", is too vague. It sounds like there's a risk of AI generated experience being ranked above crude yet enjoyable ones.

It could create a bias that culls well-meaning but inexperienced artists while boosting AI generated content that appears comparatively higher quality at first glance but falls apart when interacted with. The stance should not be "well it's solving the quality crisis even if it's AI generated"

I feel like "smart tools" needs a bit more nuance, because just because something is using a neural net doesn't mean it's using genAI. In audio there's a huge world of difference between Logic's Mastering Assistant and Suno stem generation, for example.

All I really want is better ways to report AI content. Even if the world is visually appealing, it is not worth the time of day if the creator doesn't even know of the use unity/blender lol

Any use of generative ai should be tagged. Those tags should allow people to filter it. I'm going to put my input in the forum but I really don't understand how "people want to know what's in the sauce" is something that needs clarification. Maybe different tiers of filtering but, filter it.

I'm just going to expand on this for a second too - the people arguing that we don't need to disclose the use of ai tools are often people that understand that the community they pander to is going to have backlash to it. Like Tim Sweeney complaining about steam's requirement to disclose ai use.

The "ai is used here" hasn't tanked the sales of things like arc raiders, because they were open and transparent about what was in the sauce. But it reduces sales by 80% odd as soon as you add "generative ai assets" instead of "yeah we used an ik neural net to adaptively have the enemies move"

There is a line. Peoples line varies. The important thing is you tell them what's in it. That's why we have food labels, etc. I also hope that you added the AI tag to the steam listing before dropping this post too.

World rating system, required screenshots and a tag for whether something was "human-made" would probably put the kibosh on a good chunk of these issues without the potential for abuse that tech bros are worried about. Personally would prefer an AI tag but know a major issue is that these systems

will get to a point where people won't be able to tell the difference between generated work and human-made work. Also am personally of the opinion that fully-generated content shouldn't be allowed at all but also know it's unrealistic. Rating system is based & am glad you guys are brainstorming it.

There is currently no ethical way to use AI. All of it was built through exploitation. And yes, that includes code. My code has a license for free non-commercial use. If you’re using AI that uses my code to create profit, you’re breaking the terms of my license.

I personally think 1 and 3 are /okay/, MAYBE 4. I'd hesitate to support 2 because that is something that can be open sourced, written by the community. 5, 6, and 7 should stay out.

Let’s be real. People are bad at labeling things. Have a review system that people can rate and/or comment like Steam does for games and people can read or skim. Real feedback from players.

The article says ‘AI placeholder art recently made its way to live content’, which is really telling about your approach… like, you don’t need to be using that garbage. You can just put black text on a white square saying what image will go there once it’s done. The slop step is unnecessary.

Yea... That one bothers me quite a bit. Isn't any good reason imo to just use a placeholder asset that is dubious to begin with. Just use a rectangle/box of the approximate right size with the stereotypical pink and black missing texture checkard pattern or somethin' like that.

I really hope something can be done about misleading AI world thumbnails. They have a severe impact on the once fun pastime of world hopping. After ending up in multiple slop worlds last night, we just resorted to hopping to "No AI" banner worlds.

I'd advise a full AI ban, but let users tag media they create as AI generated so that its easier for your staff to ban them. Also add a tool for reporting AI media if you find it, and block the tool from being used on things made before AI came to prominence to limit some abuse.

See image. This is in direct opposition to most of us. AI severely damages our experience, regardless of quality.

It's good to see this finally addressed, I already assumed the reason for the silence was because of this but I would like to see better communication next time, a simple "we've seen your comments and are working on a response" would have already gone a long way

I'm glad there's a response, but the team that is supposedly "struggling" to make a decision on banning AI needs to get a full reality check, no one wants AI except for people with no actual creativity in their minds. I really hope you guys take the right steps about the additional feedback.

regardless of my personal thoughts, if this was that cut and dry, we'd have a lot fewer words in that post per the graphic, people define "AI" at different levels, are worried about it in different ways, and so on i imagine most don't want slop! but people love some worlds that use ai tools

Outside of assisting those with disabilities I think AI (LLM, Agenic and Generative etc) should be banned. Classic AI can stay of course but anything that needs a datacenter for it’s brains or needs models based on collective stolen works should to be banned.

AI has multiple meanings, like mobs in RPGs for what attacks enemies do, or path finding for how units in an RTS or MOBA navigate to where players click. It's easy to see generative AI as exponential Wii Shuffleware, drowning out real hand crafted efforts, but it's also significantly worse morally.

Nuance genuinely is important here. As one example, asking an LLM tool to “make me a quest version of my avatar” is not generating any art, it’s an optimization task. I don’t think that someone who just wants their quest friends to be able to see them should have to use the “AI” avatar label.

Hard to respect your nuanced take when it involves pretending optimization, and the idea of tech art in general, is somehow lesser than the visual art. LLMs stealing from GitHub isn't inherently better than midjourney stealing from ArtStation If it's generated it's genAI

I've studied optimization for years, and am now considering a career move to tech art, I spoke with a mentor about this not even a week ago. So I care about these things quite a lot. AI image gen, to me, seems to directly displace the work of artists in a way that LLMs do not.

is it really a complicated issue? you either want a place that fosters creativity done by real people expressing themselves or you want a big dump filled with rubbish. we're all inundated with this garbage everywhere else..and you wonder why people bristle at it showing up here too?

Increasing discoverability of high quality content (user votes, user curation/listings, etc...) will solve both too-much-AI, and too-much-assets problems at once!

If any tags get added, please let us filter based on them (and most importantly, make it apply to everything, not just search, and don't make us have to select it every single time). The reason we want them is to curate our experience.

I said for people to calm down and just wait for you guys to make a statement but some people went full feral and didn't want to wait. This is progress over nothing at all being done. I'll take this over nothing. Thank you for allowing people to have a voice.

I think it's simple: Don't allow any sort of AI Generated anything within the game... Especially user generated ones. Since it'll flood the platform and devalue it.