Freesound Forums

Though this is from a US perspective, the Berne convention and US influence might make it relevant to others outside the US. It's just a lawyer discussing some issues around recent guidance from the federal government about AI authorship, registration, etc.
https://youtu.be/wy08U3gEU8w?si=hqcyDtRwPnbuzlsi

Since one of the pressing problems with gen-ai contributions to this archive is lowgrade spam -- would it be possible to put a hard limit on how many minutes-worth of contributions a user can make per week?

If someone has to think about what's worth using up their allowance to post, there'll be some built in quality control. At the least, it will be something the user *likes*, however it was made.

Hi,
I think a time or quantity limit would have to be so high that it would be useless.
None Ai uploaded, can share 50-100 sounds in a go. (Though perhaps 150 uploads in a week wouldn't hinder most users) but it might someone who did a multi-sample of an instrument.
As for time, I think some field recordist frequently upload more time wise than ai users share.
A user sharing an over hour worth of audio per day is not unusual.

I don't have specific numbers for Ai users, I'd say its in the 50 uploads per 3 days with 3min audio. So about 6h of audio per week. Currently for worst case scenario ai user.
I agree, more than your average user, but very close or less than some power-users here.
(Then those users could be whitelisted). I guess your idea could work.
Then those intensive ai users are still rare right now, often it's just a few users uploading 4-5 songs on a one-time basis. There probably less than a dozen frequent ai users here.
I presume people realise there isn't much reward in Ai content, be it street cred points or monetary or self fulfillment. Making them a short time users.

All that said, not only do I think a time/quantity limit wouldn't work without hindering some normal users, but the main AI material is music, which is rejected at the moderation process.
I see little Text To Speech, or "Ai field recording" I think, most of what goes through moderation, is a few "dinosaur screem" that some regulars here do, but I don't think that's an issue.
I might not be seeing at the right place, I'm not omnipresent on the site; feel free to share links or your side of things ^^

On another side, though the lens of Freesound as a research sound bank rather than a creative sound library, filtering for quality/usability isn't as desirable as just MORE DATA.
So I think really, all the problems and the best for the research and creative world, is to flag ai content and let users filter it in/out at their will, and rely on the rating system for quality.

Thanks for sharing your thoughts, and continuing to the discussion, looking forward to continuing it.

Im, not here to argue or debate about "AI" -- nearly 40 years in technology, I understand all of it 100% and I am 100% against it. There is nothing new you could share with me to convince me its actually a good idea.

Ive been using this site for years but only recently made a profile, donated to the org, and I came here to specifically speak against this. "AI" makes "content", it does not make music, and I will not support any org that intentionally engages with "AI" companies or individuals.

"AI" is wrong on many levels -- if you are not aware of the dangers of AI, take a look at EFF, do some opposition research (do you know what critics of AI are even saying?), and try to understand -- there is no good way for all this to end, under this economic system anyway.

Hi everyone,

Thanks for the very interesting discussion. This message is to very briefly mention that we are working on ways through which AI content can be marked as such during the sound description process (and thus filtered), and also we're working on ways through which sound uploaders can better express their preferences in relation to their sounds being used to train generative AI models (in addition to what CC licenses say). We're doing this to address both the concerns already expressed in our blog post from last June, and also concerns expressed in this forum thread.

We will let you know more about it once we have implementation details sorted out. Thanks everyone again for sharing your thoughts!

Hey freesound community, how are ya.

I've been on this site over 18 years. Oldest account in this discussion (yeah I checked).

I'm the user you other users upload for. I've been sampling yall beautiful people for almost 2 decades. Lol!

Late at night in the dark, a man sits mangling (possibly torturing) some poor innocent sound he found on this site... many many untold hours... years... there is a deep sense of respect in this process. Any maker respects the material they work with. Even a humble potter 5000 years ago respects the clay which they mold into something as simple as a cup. Does AI respect anything at all?

In other words, take solace in the fact that I'm a genuine human being. And for the record, money is never why real ones are into creative expression. Surely most of you are aware when I say, it can even us cost money (and certainly time) to pursue artistic endeavors.

Personally, I don't see any meaning in art that is not sourced from genuine expression by a real person. If I'm being brutally honest, I'd assert that it's probably the strongest form of poserdom that has ever existed. Nature can be beautiful, but nature isn't art. It has aesthetic similarities, but they aren't the same. Nature isn't expressing anything consciously. We are, though. Is AI conscious? Currently, my opinion is LOLNO. Besides, AI isn't nature. Nor is it human. It is something new, different. Likely interesting in the future.

Pick and choose your battles. Is this sword worth falling on?

Think about any convenient medium throughout history - printing press, photography, film, etc - it's not like the greatest masterpiece peak works were created the instant the technology became feasible. Peaking implies a temporal nature - these things take time. Everything is built upon one another. Culture is a layercake - none of us are here without having consumed content to cultivate our senses and sensibilities.

Being the goofball that I am, if given the choice between AI tunes or nothing at all? Today, I'd take nothing at all. But it'd be foolish or naive to think we can somehow collectively stop that cultural atomization from happening, or that it won't become more sophisticated or worthy of merit over time. And if that never appears possible in my lifetime? That's fine as well, still a lot to hopefully savor in my backlog or rehash to wax nostalgic.

If I was invested in this site from the exact opposite side - as in, had uploaded 2k samples over 18 years - and some techno-dweebs ripped all my content to throw into some monstrous poser algorithm, I'd probably be salty too.

Or... if I was invested in developing some generative content engine, I'd probably be running around like stealing everybody's content as well, no matter how seemingly irrelevant. I mean those guys are treating the internet like it's the wild west or something right now, it's bananas.

Maybe the silver lining we end up with, especially on the internet, is OG sites like Freesound become more relevant to the communal/human side of what art and connection and culture are all about. Maybe this place will become a haven, if the people who run the site navigate the future in a wise manner.

This is situation reminds me a lot of the American folklore tale about John Henry (man vs machine). Machine will win out, but respect to the people who have made this site such a great source for many different reasons, and hopefully it continues to be that way.

In closing, let me say this as simple as possible... without you, then this site just has mere sounds. Meaningless. Equivalent to AI. YOU are what makes these sounds special! It's all about the authentic human experience and AI cannot replicate that. Never forget that.

Keep it real, Freesounders. Keep it 100% human and I'll keep yoinking your goodies. Keep freeing the sound! I'll post again in another 18 years, smell ya later.

Apologies if someone else already mentioned this -- I skimmed the thread but didnt see anything: will there be a way to filter out "AI" nonsense in the future? I see the "AI" tag showing up and some people are using it, so at least for now a filter would work until people begin just refusing to tag their generated garbage.

The *energy consumption of what people call "AI" alone* will accelerate climate change emissions - already at the tipping point - by a large amount, and it has already begun. Regular energy consumers are already seeing 300%+ increases in energy bills, not because energy is harder to make but because subsidies are being given to "AI" (datacenters) companies and for-profit energy companies are making the energy. These companies RAISE RATES ON REGULAR CONSUMERS rather than build-out more infrastructure or reduce the subsidies.

Right now, the 'Taco Bell' fast food chain (Yum brands) is partnering with Nvidia to replace 10,000s of workers with "AI" (machine learning + automation, not AI). This is just one company. Many others have already begun, meanwhile customer service expectations are dropping so fast it will make your head spin. Does anyone notice the CUSTOMER SERVICE ISSUES that are rapidly growing across all industries?

How about the idea that these companies will try to replace MEDICAL PROFESSIONALS with LLMs and automation? Is anyone comfortable with *this* (western, american) healthcare system removing MORE real people and replacing them with toys???

This is an insane, obvious, hindsight-informed dead-end that anyone over the age of, I dunno, 18 can see as clearly as day. Remember the auto industry? Oh heyyy but the new auto industry that replaced the old one builds gargantuan trucks that avoid EPA regulations because they are SUVs not trucks. Do you want your nurse to be automation? Do you want your music to come from corporate brands that corner all art markets and replace all humans, de-valuing art until there is nothing left?

There are TWELVE YEAR OLDS in my children's online school classrooms being caught by the teacher on a weekly basis using nonsense LLMs to cheat on all subjects, and they are proud of it. The children doing it have parents that are allowing them to do this so they look like math, science, etc geniuses in front of other kids, and it is making teachers, scientists, etc, work hard just to figure out what papers are real!

My goodness, there are more risks than benefits to this as it all stands now! Under a different economic system - one that does not endlessly incentivize trash like this - perhaps LLM/AI would be benign or a useful tool, but as it stands now this is NOT some magic pandora's box that is now open, that we cannot close, that we must tolerate. No. No it isnt.

frederic.font wrote:
Hi everyone,

Thanks for the very interesting discussion. This message is to very briefly mention that we are working on ways through which AI content can be marked as such during the sound description process (and thus filtered), and also we're working on ways through which sound uploaders can better express their preferences in relation to their sounds being used to train generative AI models (in addition to what CC licenses say). We're doing this to address both the concerns already expressed in our blog post from last June, and also concerns expressed in this forum thread.

We will let you know more about it once we have implementation details sorted out. Thanks everyone again for sharing your thoughts!

Good to know this is still on the radar. I think maybe a blocking function might need to be introduced as well to filter out accounts that only post AI-generated content, I've already come across a couple just from browsing the homepage.

Ummm...Did I miss or misunderstand something? Now I wanted to upload my sound to the website, and I saw that they were telling me that "You haven't set your preference regarding the use of your sounds for training generative AI models." That is, there won't be a button "I don't want my sounds to participate in AI training"? Of course, I understand that for someone this question may seem rather naive...

Hi friends,
Agree with @newlocknew. It would be fair to let users choose to not participate in any AI training.

Wishing you all the best !

Kevin

Hi!

This option is a new feature we're introducing. We deployed it last week but we have not announced it yet. We're planning to do it in the coming days through a blog post.

As a result from the discussions after our "Freesound in the era of generative AI" 2024 blog post and other discussions that we've been having as a research group, we decided that we would add support for 1) better identifying uploaded sounds that were created using GenAI; 2) allowing users to express their GenAI training preferences in a more fine grained way.

Regarding 1), we added some text in the sound description form that suggests adding a specific tag if the sound was generated using GenAI.

Regarding 2), we added an option in the user profile page which gives a bit more control for expressing uploader preferences regarding the use of their sounds for training GenAI models. Details are given here, and will be soon summarised in a blog post as well, but essentially the idea is that 3 options are given, i) to follow our recommendation for the interpretation of CC licences in GenAI training context from our 2024 blog post; ii) to only allow use of sounds for training open models; and iii) to only allow the use of sounds to train open, noncommercial, models. For a reference of what "open models" mean, please check the OSI definition of Open Source AI.

There is no option to fully disallow GenAI training. After a lot of internal discussion, we decided not to include that option for several reasons, the general idea being that Freesound is a project born in a research context, and research and experimentation is part of its very core philosophy. We believe that many of the issues that AI is raising when applied to sound and music are not necessarily due to the technical nature of AI, but because of the way in which it is used, particularly in contexts where copyright and author rights do not seem to be respected, and in contexts where technology is developed with the aim of *optimising* the creative process . Nevertheless, we believe there are other ways in which AI can be used and become a useful tool for the community. As a research group that built Freesound 20 years ago to support sound sharing, research and exploration, we don't want to refuse AI, but we want to work to develop technologies to support responsible and ethical AI, to research about how AI-based tools can serve the community and enrich the creation process instead of replacing or reducing it. And this is not only about our research group: Freesound is a very important resource for the whole research community, there are more than 1000 research papers that use Freesound sounds, and it is essential for developing open sound-related technologies which are accessible and can benefit the society.

We believe that the options we provide are good to limit the uses of AI that can easily create a situation in which a third party takes advantage of the community without contributing in one way or another to it. We believe these address many of the concerns expressed by some Freesound users (and which we also heard in many other discussions that we've been part of), and at the same time promote research on open, ethical and responsible AI to benefit the community.

It is important to highlight however that AI is a fast moving field. It is unclear how to technically implement such user preferences (from a legal point of view), and there is still a clear lack of models to follow about this. We are trying to be part of the whole "AI and creativity" discussion, and we are open to make changes and adapt.

I sincerely want to believe that everything that is being done on this site deserves the highest praise and contains only positive motivation from either side. I'm really happy that I found such a place once. And I don't want to offend the creators, users, and people who are involved in any way. But behind all this jumble of phrases that seem ridiculous to me, I see that in the distance, "behind the trees," there looms such an answer: "the DISAGREE button = the DELETE ACCOUNT button." I understand that it may look different to different people. But as you've noticed, it's impossible to completely ban AI training, but encouraging outright theft is also not the best option. While AI is illegal, all these are just attempts to cover up with a fig leaf. It's like building a house without a foundation. Maybe it will stand somehow. Except, no. The laws of gravity cannot be defeated. A reasonable society should be different. But it's the money that decides. And this is the way to a bad ending. And the creators and authors suffer more, as you have noticed. At their expense, someone always lives happily, and the authors are content with bread and salt. And your innovations do not solve anything and there is no transparency in them. And all this is very similar to another tacitly imposed sentence that says – accept it, you have no other way. So, apparently, you still have to think about whether to press the very button. Without negativity and anger.

I think that, by now, most people interested in art as music/sound can see and understand the GIGANTIC risk to us all that "AI" (its not artificial intelligence, for one, its LLM / advanced automation using stolen training data) represents. Some time has passed since this thread was started, and Im sure *most* of you have seen real news about what is actually being done with these technologies.

1) insanely unsustainable energy consumption, across the board, with "AI" companies openly violating all emissions and resource consumption laws with the help of fascist state officials who are openly, proudly, LOUDLY corrupt

2) these systems are OPENLY planned as a REPLACEMENT for millions (read: 60-80% of some industries) jobs. There ARE NO CONTINGENCY PLANS to get people new jobs. Does anyone understand this? This is coming for ANY industry sector you allow them into (allow by eagerly handing them all your data and permission to use it)

3) numerous studies show HUMAN COGNITION is at risk: when someone uses "AI" slop for a few months to accomplish a shortcut to knowledge, they LOSE THE ABILITY to achieve that knowledge on their own.

4) the knowledge these systems contain, described by some as "oracular", is POISONED knowledge: it is biased at a fundamental level towards wealthy, upper class, white, male, western ideologues and reliably comes up with content that reflects this. Additionally, it is possible to make these models give up confidential, private, or copyrighted materials they were trained on.

5) these models all require ONGOING TOTAL SOCIAL MEDIA SURVEILLANCE, every single day, every single year, to continue being even somewhat relevant. As people stop using these platforms, "AI" models begin eating themselves, training themselves on more and more of their OWN DATA.

6) fascist governments are OPENLY bragging about using these products to a) de-anonymize people and activists, b) violate human rights including murdering innocent people autonomously, c) using these systems to psychologically manipulate children and people who use social media, d) using these systems to create a dystopian, easily hacked ALPR surveillance system that violates at least the US constitution and bill of rights

Etc etc. There are more terrible outcomes for this technology that I dont even have the patience or brainpower to write about today, but I KNOW that most of us, by now, can see the rising threat from this around the world, especially inside the US. Imagining that this is "inevitable" and that there is "no way to stop it" sounds like a culture that has given up and has resigned itself to not caring about life -- I know my family and I will never stop fighting this terrible technology, but how long will it take yours? Maybe once you lose your job, will it motivate you? Or does it take more than losing income, homes, healthcare, etc etc? Is there something that will do it?

I dont see a lot of folks talking about big studies and conclusions on this tech, while holding strong (read: supportive) opinions about said tech.

Dont take my word for it -- maybe take MIT's word for it.

https://www.media.mit.edu/publications/your-brain-on-chatgpt/

There are DEFICITS occurring in people who use these technologies. And this is on top of the other major ethical, consumption, etc issues that already exist. This is not some fun new toy to experiment with (its just research, you see!). This is a major unsustainable development on TOP OF many other unsustainable technologies. This is much, much bigger than... art and sound design, frankly.

Hi friends,
Deeply agree with the recent posts from @Milkman1976 and @newlocknew.
Of course, I can understand @frederic.font’s sight, and I encourage the valuable efforts of Freesound to build transparent relationship with respectful partners.
However, let’s be honest, if Freesound doesn’t add any audio watermark in the playable or downloadable files, will it prevent any of the numerous bots raving on this website everyday to use our sounds to train any kind of AI (open source or not, commercial or not) ?
And, given the current situation, clicking on an ‘’express your choice’’ button would not change anything to the deal, I’m afraid...

Wishing you all the best !

Kevin

Any audio enthusiast familiar with live sound reinforcement will understand the concept of positive feedack - howlback!
Well as LLM AI starts feeding more and more on it's own self-generated content, that's exactly where we're going. It's going to leave the rails and start steamng of in it's own bizarre direction.

I hope this happens suddenly enough for all of us to notice before we turn into slowly boiled frogs.

Use it or lose it. Applies to our brains as much as to our bodies.
Why bother thinking about things, remembering things, putting effort into creating things, developing your mind, when all "answers" are instantly accessible from the little glowing, speaking slab in your pocket.
It won't take generations of Darwinism for Human intellect to decline. It can happen in an AI induced spiral in a a single generation for large swathes of the human population.

Old school rant over.

Yes! Guys, you're damn right. This is all some kind of unbearable madness!