Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Open-sourcing AudioCraft: Generative AI for audio

ai.meta.com

Post reply on HN
Print view View original

321–330 of 335 posts

Re: Open-sourcing AudioCraft: Generative AI for audio

#321

Post by pbhjpbhj »

kmeisthax wrote: post #275
pbhjpbhj wrote: post #268

Earlier quoted context omitted.

>If you just want to use something at home privately you don’t need a CC license… // I presume you mean in USA, because in UK you don't have a general private right to copy. Our "Fair Dealing" is super restrictive compared to Fair Use.

Funnily enough in the UK they actually tried to fix this. The music industry argued that the lack of a private copying levy made legalized CD ripping into government confiscation of copyright ownership... somehow. The UK courts bought this, so now the UK government is constitutionally mandated to ban CD ripping, which is absolutely stupid.

I knew CD ripping got reversed but not the arguments against it, definitely stupid as not giving a monopoly is not the same as confiscation (seems like a very straightforward reasoning). No doubt done Tory got a 'management consultancy' gig with the RIAA from that one.

I like that it makes software like iTunes contributory infringers for enabling mass copyright infringement.

Re: Open-sourcing AudioCraft: Generative AI for audio

#322

Post by crazygringo »

ysleepy wrote: post #297
crazygringo wrote: post #276

Earlier quoted context omitted.

> and reporters are not interested in reporting them You really think that if the NYT started altering its past stories, other publications would just... ignore it? It would be a front-page scandal that the WaPo would be delighted to report on. As well as a hundred other news publications. Thankfully.

That is maybe true for a small percentage of stories. You are also reducing this argument to the most construed straw man instead of engaging with the idea in earnest. If you can't alter world news headlines, you can still alter the tone of the article. If you can't alter front page news, you still can alter the remaining 95% of news. Influencing public opinion is more subtle than the one important headline per day.…

You seem to be misunderstanding.

This isn't about a small percentage of stories, it's not about tone, it's the fact that if the NYT ever did this even once with the intention to truly "alter the past" it would be a major scandal.

And obviously things like corrections or taking down libelous content aren't included.

So no, I'm not constructing any kind of straw man here. I'm saying that the threat of subtly nefariously "altering the past" isn't realistic because it would be caught and exposed and there's no financial motivation to do it in the first place.

Re: Open-sourcing AudioCraft: Generative AI for audio

#323

Post by __loam »

polygamous_bat wrote: post #304
__loam wrote: post #210

Earlier quoted context omitted.

I would be pretty shocked if meta were that far ahead of all 3 of those companies, all of which are also spending a fuck load on internal AI research.

> all of which are also spending a fuck load on internal AI research. But their internal research stays internal. Sometimes, they put out "papers" which are glorified advertisements, often going as far as hiding the model architecture just to keep their competitive advantage.

I get that, I'm just saying the original statement, that meta is further along than all of those companies combined, is a pretty wild claim.

Re: Open-sourcing AudioCraft: Generative AI for audio

#324

Post by Somteencead »

The Encodec project's focus on AI-generated sound effects offers practical applications across various industries, from gaming to film production. It's exciting to think about the possibilities of AI enhancing audio design and creating immersive auditory experiences.

Re: Open-sourcing AudioCraft: Generative AI for audio

#325

Post by cwkoss »

chefandy wrote: post #320
cwkoss wrote: post #319

Earlier quoted context omitted.

Nobody is necessarily owed money for creating art, regardless of whether it was done by hand or with machines (unless they were specifically commissioned to create the work). The economics of art are subject to market forces, and they always have been - long before AI became an additional factor (among many others) in why an artist may not earn a living from their art. Deviantart in the 00s was a huge repository of a…

Great non-answer. None of the people whose art was sucked up by these machines had any idea that this would be integrated into for-profit tools and used against them in the market place, and almost certainly would not have consented if they had known that. The ones with copyrighted images didn't even give legal consent. The fact that Getty's logo showed up in red carpet images is a symptom of a problem that obviously…

The vast majority of the world's population doesn't have the privilege of going to art schools and training for years to render their ideas.

AI allows more people to be artists without the skill barrier, and that is a social good.

I'm sorry that the mechanistic portions of your art career are rapidly losing economic value, but I think free tools like StableDiffusion (and the better tools that are coming in the future) should be available to every child (and adult) in the world. And the world will be a better place as a result.

Re: Open-sourcing AudioCraft: Generative AI for audio

#326

Post by chefandy »

cwkoss wrote: post #325
chefandy wrote: post #320

Earlier quoted context omitted.

Great non-answer. None of the people whose art was sucked up by these machines had any idea that this would be integrated into for-profit tools and used against them in the market place, and almost certainly would not have consented if they had known that. The ones with copyrighted images didn't even give legal consent. The fact that Getty's logo showed up in red carpet images is a symptom of a problem that obviously…

The vast majority of the world's population doesn't have the privilege of going to art schools and training for years to render their ideas. AI allows more people to be artists without the skill barrier, and that is a social good. I'm sorry that the mechanistic portions of your art career are rapidly losing economic value, but I think free tools like StableDiffusion (and the better tools that are coming in the future…

Haha... Nice ham-fisted non-sequiturial attempt to make the artists the privileged ones here. Who can and can't go to at school is irrelevant. My example clearly showed that you're understanding of the "economy of art" is not even close to representative. Commercial artists operate at all levels of society, all over the world, and have for a really long time. Nobody making beautiful hand-carved and painted signs for local businesses started out as an independent fine artist using signs as a medium, hoping to get their big break. They learned the arts of carvin, painting, lettering, illustration, and so forth, and the craft of making and finishing durable outdoor wooden structures, and started a career.

The utilitarian argument is only philosophically defensible in the street car scenario; when people are deliberately pulling the levers unprompted and it could have been done ethically but it wasn't because it was just to darned inconvenient and/or expensive, the greater good argument doesn't work. It's the same argument people used to defend the Tuskegee experiment and its ilk... And roman public slaves. If we're willingly throwing people into the spinning wheels of progress because it will be wonderfully convenient and neat for non-artists to use other people's skills instead, there is just no ethical defense. Knocking the stool of from under someone using a tool they built and you didn't even have permission to use is just not ethical. You and many others end up falling back on right-wing platitudes about the free market.

I'm actually a technical artist— my skills are now dramatically more valuable. That doesn't make it any more ethical even if it works on my favor.

I was a developer for 10 years, and have seen this unadulterated hubris in nearly every group of developers I've ever encountered— when you find yourself explaining someone's job and industry to them, you might want to stop and ask yourself... "Am I pulling a Dunning-Krueger right now?"

Re: Open-sourcing AudioCraft: Generative AI for audio

#327

Post by ysleepy »

crazygringo wrote: post #322
ysleepy wrote: post #297

Earlier quoted context omitted.

That is maybe true for a small percentage of stories. You are also reducing this argument to the most construed straw man instead of engaging with the idea in earnest. If you can't alter world news headlines, you can still alter the tone of the article. If you can't alter front page news, you still can alter the remaining 95% of news. Influencing public opinion is more subtle than the one important headline per day.…

You seem to be misunderstanding. This isn't about a small percentage of stories, it's not about tone, it's the fact that if the NYT ever did this even once with the intention to truly "alter the past" it would be a major scandal. And obviously things like corrections or taking down libelous content aren't included. So no, I'm not constructing any kind of straw man here. I'm saying that the threat of subtly nefariousl…

You are misunderstanding. Altering the tone, comments, bias etc. is altering the past, this is what I meant originally. You came up with the straw man of the new york times. Sure, some high visibility publications are less able to be altered, but that doesn't mean none can in no way.

Re: Open-sourcing AudioCraft: Generative AI for audio

#328

Post by cwkoss »

chefandy wrote: post #326
cwkoss wrote: post #325

Earlier quoted context omitted.

The vast majority of the world's population doesn't have the privilege of going to art schools and training for years to render their ideas. AI allows more people to be artists without the skill barrier, and that is a social good. I'm sorry that the mechanistic portions of your art career are rapidly losing economic value, but I think free tools like StableDiffusion (and the better tools that are coming in the future…

Haha... Nice ham-fisted non-sequiturial attempt to make the artists the privileged ones here. Who can and can't go to at school is irrelevant. My example clearly showed that you're understanding of the "economy of art" is not even close to representative. Commercial artists operate at all levels of society, all over the world, and have for a really long time. Nobody making beautiful hand-carved and painted signs for…

> non-artists to use other people's skills instead

Here is where I think you're wrong: Everyone is a potential artist.

Many never have the opportunity because of economic reasons: some people don't have the time to cultivate the mechanical skills necessary to render their vision.

All artists pull from the work of others constantly. That is 'using other people's skills'. It is absolutely privileged to have the opportunity to devote time cultivating art production skills (which is usually done by studying and replicating the process of the artists that came before them).

If someone is born to a single mother, grows up having to raise their siblings, then has to immediately start working to help support their family, they don't have the privilege or time to explore art production skills. Technology has been making art production more accessible for hundreds of years, and that's a clear net good. I want everyone to be able to make art if they want to.

It up to the artist if they want to manifest their vision without creating the pigments and brushes themselves, or using a drawing tablet, or using generative AI tools - it all can be art that is meaningful to the person who creates it and more art in the world is a good thing.

Commercial art isn't a very good representation of the artists soul: they are constrained by market forces or their patron rather than producing what their own heart desires. We should all be happy to sacrifice commercial art jobs if it means that every person in the world gains the ability to render the art that comes from their soul.

Re: Open-sourcing AudioCraft: Generative AI for audio

#329

Post by chrisjj »

emporas wrote: post #287
chrisjj wrote: post #285

Earlier quoted context omitted.

Bit is that "extension" to 5mkns more than just the repeat of e.g. 15secs heard in the demos?

Yes, of course it is more than that. You can create full songs of several minutes. I have published 20 songs on youtube of 2-3 minutes each. Listen to that song, the best i have made so far [1]. It doesn't repeat at all! I haven't looked closely to this release, is the audiocraft page on github different than facebookresearch/audiocraft? The other two modalities may be new, AudioGen, and EnCodec, but i was under the…

Thanks. What I hear there is 19s of material on repeat, with the parallel parts variously brought in and out, and no other variation except in decorative detail.

So yes not a literal repeat, but no not enough to rate as a "full song", IMO.

Plus it is not actually song at all, note. Singing would particularly expose its repeat-and-vary trick.

Re: Open-sourcing AudioCraft: Generative AI for audio

#330

Post by OrangeMusic »

xdennis wrote: post #207
hackernewds wrote: post #137

Earlier quoted context omitted.

who gets to declare what is the "open source definition" and why?

The people who created the term: the Open Source Initiative. Before, people most often used "free software" as defined by the free software movement, but some disliked this term because it's confusing (most think "free" means no money) and perceived to be anti-commercial. The term "open source software" was chosen and given a precise definition. It's dishonest, then, for people to use the term "open source software"…

> It's dishonest, then, for people to use the term "open source software" with a different interpretation

I disagree. You're saying that they "invented" the term, but it's a very generic term. The source is open, it's open source. I bet people were using the term before they claim they invented it.

In that context, it is very fine to use a different definition and in fact here's my definition, and I guess most people (maybe not on HN) share it: if the source is visible by the general public, it's open source.

For what you mean, I use "FLOSS".

Post reply on HN