George Carlin’s heirs sue comedy podcast over “AI-generated” impression

Suit alleges copyright infringement and illegal use of Carlin's name and likeness.

See full article...

 

[IncorrigibleTroll]

I get where you're coming from on the Black Mirror episode. Personally, I thought it was one of the best in the series. It explores the idea of 'resurrecting' someone using the data they've left behind, which, while it sounds like sci-fi, isn't entirely out of the realm of possibility with advancing technology.

Ray Kurzweil, has even talked about a similar goal to 'resurrect' his father using collected data. This idea sounded far-fetched to many back in 2010 (It only took me a short time to 'get it' though), but as technology evolves, the concept becomes more conceivable. It's like how .par files work in data recovery. They reconstruct missing parts of a file from the remaining data. Similarly, with enough data about a person, theoretically, you could recreate a semblance of them.

While it's not exactly the same (especially recreating their experiences), this concept is fascinating. It shows the potential of technology to bridge gaps we once thought unbridgeable. Sure, it's a complex and sensitive topic, but it's undeniably intriguing to think about the possibilities.

Ray Kurzweil has never really been able to deal well with the concept of mortality, has he? He's been obsessed with radical life extension since long before it became a techbro thing.

Sorry Ray, you're still going to die someday. We all will.

 

[IncorrigibleTroll]

IMO? Cause it's better and may someday be indistinguishable.

Ars posters have drawn a line about jobs AI can take and it seems Them Pictures and arts stuffs are the line that has been drawn. Everyone else can be outta a job but them arteests can't be touched.

It'll come for all our jobs eventually (sure, people can laugh all they like with all the mistakes and hallucinations and limitations of the current tech, but they won't be laughing in a few years) and anyone who thinks there's any stopping it is very likely in for an unpleasant shock. "Information wants to be free" isn't an ethos; it's an observation about putting toothpaste back in the tube or keeping Pandora's Box closed. We'd be better served figuring out how what the impact is going to be, how we can accentuate the benefits, and how we can mitigate the harms.

 

[IncorrigibleTroll]

Sadly, that's the truth. This OP is about an hour long presentation of an AI trained to sound like George Carlin and which does so while being slightly, but noticeably off.

Tomorrow's OP will be about russians and republicans crafting similar disinformation on social media in order to make it look like their political and military adversaries are monsters, clowns, or both.

And you just know that the third of the US voting population which went all-in on Covid parties and Ivermectin to prove their loyalty to Dear Leader are going to eat that shit up.

I'm not sure that crowd really needs gAI to buy into nonsense. They want to believe, so a crudely scribbled napkin is likely sufficient evidence.

 

[IncorrigibleTroll]

Yeah, the iconic (alas) dumb surprised face on a Youtube thumbnail is a good indicator that I can block the channel and never watch any of the shit they do.

So you're saying you really only listen to music or watch clips of TV shows on Youtube? Because that dumb surprised face is omnipresent. Honestly, it should really be the Youtube logo at this point.

 

[IncorrigibleTroll]

It's mostly Stephanie Sterling, Yong Yea and some local creators that are not dumb

A friend of mine has a little niche channel about 90s/00s drum & bass records. He hates the stupid face thumbnails. But the viewer numbers don't lie: his videos do markedly better with a stupid face thumbnail than without.

I don't think we can put much of the blame for this on the youtube creators. They're only responding to the behavior of the viewers as deduced by the Great and All-Knowing Algorithm.

 

[IncorrigibleTroll]

Of course. Now consider the future where you want to find a given youtube clip of 'politician A' who said something worth hearing but, since that spiel was about gun control the NRA has flooded youtube with bit reels where he advocates the use of baby blood for rejuvenation, or confesses to being aroused by young children, et cetera.

Not to mention all the flicks we can expect where some person conspicuously identifiable as part of a given minority group, 'accidentally' spills the story of their organized part in, oh, grooming, Great Replacement, kidnapping white women...you get it.

What AI can do is to flood the market with bullshit you can't separate from the real quickly. Maybe that'll just screw youtube over, but at least for a time I predict we'll be seeing a whole lot of shit like this.

I very much suspect that the opposite scenario is far more likely, at least longer term. You can really only dupe most people with generated imagery a couple of times at most. I think it's more likely that people will stop trusting in any imagery than it is that people can be fooled repeatedly by generated imagery. We've been seeing that on a small scale for years now, where people will argue that a particular image is photoshopped if it's unfavorable to their position. Generative AI increases the speed and scale of an existing phenomenon (convincingly altered images), so I expect it will likewise increase the scale of the existing reaction to that phenomenon.

 

[IncorrigibleTroll]

I get the temptation not to let a troll get away with trolling, but every quote surfaces the message. I recommend ignoring and/or downvoting.

Whatever else one might say about Kamus, he's clearly a true-believer. He isn't trolling, he's evangelizing.

 

[IncorrigibleTroll]

It's the "other than" part - "other than distribute copies or derived works."

First sale doctrine says that if you buy a CD of Metallica music you can sell that CD to someone else. It doesn't let you sell copies of the CD to someone else, nor does it let you rip individual tracks off and sell them to someone else.

What's more it also doesn't allow you to take those tracks and transform them into a derived work - like by incorporating them as a sample into a work you're creating. (which is why samples are licensed works).

Generative AI is all about sampling and creating derived works. The question that is still open is whether it's being done in a way that is going to be considered "fair use" or if it's being done in a way that ultimately model makers are going to need to be paying for. The same way that musicians have to license samples to create new works.

The sampling regime as it currently stands is rank bullshit that stifles creativity in service of capitalism and is not something to model in other areas. Toxic rentseeking power games.

 

[IncorrigibleTroll]

Why do you feel you're entitled to use the work of others without their permission?

Because it's inherently transformative (assuming you put in the work to properly flip it. I'll never stand in defense of laziness like Ice Ice Baby). Because masterpieces like It Takes A Nation of Millions or Paul's Boutique are basically impossible now. Because some of the most original and unique artists I've ever heard sample heavily. Because commercial considerations must always be weighed as secondary to creative expression. Because using samples imposes certain interesting creative problems to solve. Because copyright is a form of rent and rent is unproductive deadweight even by capitalism's own precepts. Because ultimately all sound is potential grist for the creative mill. Because I do not accept that ephemera can be owned. Because I don't give a rat's ass about anyone's ability to commercially exploit creative endeavors. Because I find copyright morally and ethically despicable.

I can keep going like this for pages.

 

[IncorrigibleTroll]

It really isn't.

It really is. You're clearly out of your depth here. I mean, if we're just going to make assertions, I can back mine up with demonstrations and you can just pout.

Bullshit. You're just unwilling to put in the work.

Again, this is your ignorance talking. Soundclash doesn't (cannot, even) happen through a single desk. And sampling involves shitloads of work. You know who just slaps a bunch of samples together and calls it a day? Teenagers who just grabbed their first cracked copy of FL Studio. Anybody worth listening to is going to be chopping, processing, recoloring, timestretching, repitching, etc the hell out of their samples, because that's where the artistry and the pleasure both live.

Unless by "put in the work", you mean the work of sample clearance. In that case, yeah, sure, I'm not willing to wade through that nightmare. Maybe look a little more deeply into what's involved with that. It's a disastrous shitshow of a process with more veto points than congress, and it's kind of adorably naive that you think any more than a nominal fraction of any fees would make it to the actual artist. Sampling is literally the worst fucking model to cite. Cover songs (which are not the least bit transformative and much less creative than even the laziest, most adolescent sampling) have an infinitely better licensing model than samples (cumpulsory, fixed rate, percentage of revenues) and I wouldn't be so incensed had you pointed at that.

They're not original or unique if they rely on other people's work that heavily.

Really now, this is nothing more than a childish circular wank, not an argument. I'd rather not engage it at all, but if you're trying to say that Death Grips sound like Jane's Addiction because of a few samples, you should probably just close your mouth on this topic forever. Dadaist poetry also "samples" heavily, and that's clearly distinct from other styles poetry or prose. You mostly remind me of those asshat guitarists who used to troll the synthesizer forums on gearslutz to dictate what is and is not "real music", so I'll rate your opinion about as highly as theirs: somewhere well below the sub-basement. Good for you, you know 3 chords. We're all so very proud of you. Now go away.

Nobody gives a shit how you feel. You just want to be able to take from others without compensating them for it.

This argument cuts both ways, bub. I don't give a shit about how you feel about it. I've never cleared a sample and I never will, and I feel not one iota of guilt. Nor would I be the least bit offended or bothered if somebody samples one of my tunes (though I would love a heads-up so that I can check out how it was used), and I most certainly would not be sticking my hand out about it. Get your fucking toxic moneygrubbing out of my art. IP maximalists are just landlord apologists in a different context, and I have equal sympathy for both.

Then why should anyone give a rat's ass about your ability to use other people's work because you can't come up with your own stuff?

And again, the chump who knows fuckall about that of which he speaks makes hard and fast declarations. Fuck off with the condescending nonsense, you ass. Sampling a rottweiler bark from somebody's youtube video to use as one layer of several in a snare drum is coming up with my own stuff. If you're going to insist otherwise, well I don't know what to tell you other than that you're speaking out of your asshole (and if the timbre is interesting, I might sample it and use it in the mids layer of a bass). And I don't owe that person any more than a kidnapper owes the magazines they cut the ransom note letters out of.

Again, nobody cares how you believe that you're entitled to the work of others for free, and that artists don't deserve compensation.
You don't have to, we all know your only reason for this is that you don't believe artists deserve compensation for their hard work.

You speak for everyone, do you? And don't make ignorant statements about what I do or don't pay for. I can just about guarantee I put more money in the hands of artists than you do. But by all means, continue to feel incredibly smug because you give spotify $15/month.

 

[IncorrigibleTroll]

It helps to remember that Kamus writes most long form posts via LLM. The LLM doesn’t take rebuttals into account. It’ll never “learn” from discussions here.

Though the LLM he uses may not be the only one with that problem.

The tech is impressive, but it isn't quite good enough for that usage yet. That does go a long way to explaining why even on the occasions when Kamus makes a point I'm inclined to agree with, he makes it in a really lousy way.

You're a lawyer or at least in a law-adjacent field, right? Those dolts submitting briefs written with ChatGPT must be a great source of amusement and professional facepalming. It's good enough for fluffy business copy (the stuff that was a lot of words to say little to begin with), but not much else yet.

 

[IncorrigibleTroll]

The reason you 'sometimes agree with me' stems from the fact that cynicism often struggles to stand up against this thing called 'reality'. Consider both of your stances on Bitcoin: despite numerous pronouncements of its demise by people like you, it remains resilient and relevant. Trends backed up by evidence always have the final say.

Vindication may take time, but when the tide turns, even the staunchest skeptics, like yourself, might find themselves adjusting their stance. Hell, It wouldn't be surprising if, by the end of this year, you finally capitulate and buy some Bitcoin, possibly through an ETF. (which is just an IOU, and kind of defeats the purpose of getting into Bitcoin in the first place, but whatever )

Regarding the functioning of LLMs and their ability to 'remember' rebuttals: While LLMs don't have memory in the human sense, they operate within context windows. This means if a rebuttal or any piece of information is within the current discussion's context window, the LLM can access and use it for generating responses. It's not about recalling past conversations but about processing the available information within the current interaction's scope. This method allows for coherent and contextually relevant responses, as long as the discussion details remain within the LLM's accessible context. So, while It can't 'remember' past sessions, It can maintain continuity and address points effectively within an ongoing conversation.

Moreover, the actual reason I often find myself repeating points, especially to individuals like hillspuck, isn't a shortcoming of the LLM's memory capabilities. Rather, it's a necessity to continuously address and counter persistent skepticism. When someone, like him admits to not fully engaging with detailed responses, it becomes inevitable that my points need reiterating.

TL;DR.