This is a most excellent place for technology news and articles.
This is probably so that they can create translated versions of them, so if your audiobook is only in English and you upload it you can check a box to have it also be available in other languages you'd never have been serving otherwise.
It's almost certainly expanding on the same service they added for podcasters:
(A translation is a derivative work.)
Likely. They want something for nothing - free translation without paying a translator, licensing an official translation, paying a voice actor, etc. If the TOS only said that it would already be extremely problematic.
In fact the language is so much more broad than that.
I mean, at a certain point this kind of thinking becomes like the MPAA's math around thinking every person downloading a movie from a streaming service was a lost sale.
Yes, this would mean a massive expansion of translated audiobooks without the labor that traditionally would have gone into creating them.
But we don't have translations for the majority of audiobooks in the majority of languages because the costs of that labor has historically outweighed the benefits of a potential expanded audience in niche languages for the long tail of audiobooks.
Personally, I'd rather live in a world where there's broad accessibility to information for all people regardless of their native languages, rather than one in which humanity tears down its own tower of Babel to artificially preserve the status quo.
Ah yes the "labour should be free" / "but if we have to get permits from every artist we won't be able to feed our AIs!" argument.
Listen, I'm not gonna lie. it'd be wonderful if we lived in the utopia where everything is autotranslated for us (not to mention it's done correctly, no "Brock's jelly donuts"). But there's 123456 ways to get it done with human labour properly paid and the corporations are in the position where they have the power and the responsibility to do it. Else authors are going to end up with automated translations which are sold as "official" but over which they don't have control, in particular if the AI translation misrepresents them (using language the author wouldn't changing concepts, or even - imagine - adding slurs).
Like, sure, maybe these corpos don't want to pay for someone to do the translation from scratch... but have they thought of looking for fandom translations and sourcing and paying for those? That's work already done, and has the advantage that someone cared enough about the "niche work", kinda like with anime fansubs. Or they could also, you know, novel idea and all, pay people a wage to translate this. I know. The horror. How dare I suggest that a company doesn't divert wages and income to the CEOs!
I hope that once enough people get replaced with automation, they'll realize how shit capitalism is and push for harsher corporate tax to fund UBI.
That's fair, and I have no problem with authors employing machine translation in order to translate their works. However, I happen to think that that should be the writer's decision.
Most authors would much rather employ a professional translator to get it right instead of a computer to approximate it. He
However, I happen to think that that should be the writer's decision.
I don't know why you think it won't be.
What, you think Spotify is just going to do it without the uploader choosing whether the feature is turned on or not?
The podcast translations are opt-in. Why do you think these won't be the same thing?
Because I wasnt born yesterday, and dont spend my time with my head lodged firmly up my own asshole?
Look at how broad that legalese is. Ask yourself if auto application would make them more money. Now count what 2 + 2 is on your fingers.
Maybe they just want to include clips of the audio book in user’s yearly review thing.
That's plausible and I'm a little rusty on my IP here but I would call that a fair use. Derivative works use existing work in a new way, where the added creativity is sufficient to make the new work itself copyrightable.
Fair use doesn't exist in all countries though.
This will be an unpopular opinion here.
I'm not against AI but the rules have to be in laws and regulations. First, AI can't use copyrighted material without paying for it. It can't either use material without asking individually.
The second point is that AI can't created copyrighted material. Whatever an AI created, it's free of copyright and everyone can use it.
Third, an AI can't be a blackbox. It has to be comprehensive how it works and what the AI is doing. A solution would be to have source available code.
Fourth, AI can't violate laws, create and push misinformation, and material used for misinforming.
And, of course, anything created using AI has to be indentified as such.
The money is in what the AI can do, the quality of the result, and the quality of the code. All the other things isn't valuable.
I'm fine with this as long as the "pay & ask" has an exception for non-commercial, open source projects, otherwise it would mean that only corpos can create models, and everyone else is SOL and thoroughly fucked, because they will pay a license fee to the platforms, and the platforms will just add a new TOS element that by using the platform you consent and withdraw your rights to compensation.
I imagine that if AI devs didn't sneak around copying people's works in bulk but instead asked for permission or paid for a license, artists wouldn't hate it like they do now.
Yeah artists were fine with publishing companies doing this since the dawn of literacy but this time it is completely different
My gut feeling says that's not entirely true. Generative AI has so many qualities that make could it offensive to so many people, I think we were going to see a pushback from artists regardless. The devs' shitty training practices just happened to give the artists a particularly strong case for grievances.
Your third point is an active research topic, we can’t explain exactly what generative (and other) models do beyond their generic operation.
If we could explain it, it would just be another rules engine 😅
Most of those laws are unenforcable and some are even undetectable.
Your ideology is getting in the way of objective fact.
1 & 2 are... #3 is impossible, though...
Are you kidding? #3 is the second most possible one of that set, it's just a matter of setting up Reproducible / Deterministic Builds.
If you can't replicate a result with control of the software version + the arts input + the randomness seed, then "something else is going on".
deterministic builds?
the "builds" in ai are 1,000's of hours of supercomputers randomly mutating and evolving a gigantic neural network...
the inner workings of such are very much a black box.
to try to save that in a perfectly reproducible way is completely unreasonable, and simply will never happen.
you could require all of the arts input to be documented and saved, but people would lie and you're talking about a very large amount of data being saved for however long... also not really reasonable...
and you also have to understand that there's a lot of countries in the world, computers are all connected on the internet, and ai will just run in other countries, and illegal systems would run in the whatever country is dumb enough to try to but completely unreasonable and expensive extra requirements like that on it.
there's a whole field of study trying to reverse engineer neural networks after they're created... i.e. it's a black box to the people that make it
The only way to make a clear text LLM is to convert most of the hard storage that humanity produces for the next ten years into storage, and we'd need about 1/4 the processing power of bitcoin mining to have it run at ChatGPT speeds.
Even said, blackbox self-modifying AIs will be the models that win the usefulness wars, and if one country outlaws them then the only result is they will have no defense against countries that don't feel the need to comply with them.
so, your first paragraph isn't true. but i'll point out that bitcoin is mined with ASIC chips entirely now, which only hash bitcoin transactions... they can't compute anything else so it's not really comparable...
second part i do agree with except for self-modifying... although that doesn't seem too far away...
You really don't understand how LLM data blobs are created, do you? Nor do you understand how ridiculously compressed it is?
what's does that have to do with anything?
