Sarah Silverman leads class-action lawsuit against ChatGPT creator

[J.Quondam]

Help make folks who only read headlines angry?

Well, for the benefit of all the headline-only readers out there, the linked article article notes:

— The chat aspect is important: There is no audio for now, although that is apparently coming next year. I can’t help but wonder: Will it be more successful than Amazon’s celebrity Alexa voices, which it shut down earlier this year?​

​

— Absolutely no detail on the business model behind this. We asked how Meta was compensating the celebrities for their likenesses, which got a resounding “no comment” from the spokespeople. ​

So it seems reasonable to say the objective of this is still somewhat murky, at best.

(eta) Parmandur posted this as its own topic, so further chatter about it shoudl probably go over there:

https://www.enworld.org/threads/meta-shows-off-an-ai-dm-played-by-snoop-dog.700133/#post-9143625

 

[overgeeked]

 

[trappedslider]

Tom Hanks Warns of an AI Version of Him Used to Promote Dental Plan: “I Have Nothing to Do With It”

Tom Hanks is warning fans not to trust a video of him circulating about a dental plan. The Oscar-winning actor took to his Instagram on Sunday to give his fans a heads-up that anything they may see about him online linked to a dental plan was not actually him. “BEWARE!! There’s a video out there …

www.yahoo.com

 

[Scribe]

As if the 2020's havent been a clown show enough already.

 

[Reynard]

View attachment 296633

Look, I absolutely believe that copyrighted works should be opt in for training AI, but I am constantly cringing at how poorly people understand the methodology and what happens in the algorithm. Especially, but not limited to, older creators.

It isn't a collage machine. It doesn't take chunks of your art or text or code and insert it into a new thing. That is not how it works. It is a blind algorithm trained to decide what follows the previous thing based on specified prompts. That could be an adjacent pixel or the next word in s sentence or a following line of code.

The reason you see near or exact reproductions is that some sources are pervasive in the dataset. The most popular example is the second paragraph of Harry Potter and the Sorcerer's (Philosopher's) Stone: it shows up because a HUGE number of reviews of that book include the first couple paragraphs. Put in the 756th paragraph and you won't get anything like the 757th.

I think it is important to protect intellectual property. And I think it is important to support, pay and uplift artists of all kinds. But I also think this kind of generative AI is ultimately a tool akin to the camera and will, in the end, create far more human developed art than it destroys.

 

[overgeeked]

Look, I absolutely believe that copyrighted works should be opt in for training AI, but I am constantly cringing at how poorly people understand the methodology and what happens in the algorithm. Especially, but not limited to, older creators.

It isn't a collage machine. It doesn't take chunks of your art or text or code and insert it into a new thing. That is not how it works. It is a blind algorithm trained to decide what follows the previous thing based on specified prompts. That could be an adjacent pixel or the next word in s sentence or a following line of code.

The reason you see near or exact reproductions is that some sources are pervasive in the dataset. The most popular example is the second paragraph of Harry Potter and the Sorcerer's (Philosopher's) Stone: it shows up because a HUGE number of reviews of that book include the first couple paragraphs. Put in the 756th paragraph and you won't get anything like the 757th.

You were wrong about this the last time and you’re still wrong about it now.

 

[Reynard]

You were wrong about this the last time and you’re still wrong about it now.

Look, you can assert whatever you like. But denial isn't going to change anything.

 

[overgeeked]

Look, you can assert whatever you like. But denial isn't going to change anything.

Exactly. So stop denying this “AI” is a plagiarism engine. We’ve been through this. You can literally get whole chunks of text verbatim from these things. As shown earlier in this thread and the others like it.

 

[Reynard]

Exactly. So stop denying this “AI” is a plagiarism engine. We’ve been through this. You can literally get whole chunks of text verbatim from these things. As shown earlier in this thread and the others like it.

I'm not going to argue with you. If you refuse to understand how the technology works that's on you.

 

[trappedslider]

Exactly. So stop denying this “AI” is a plagiarism engine. We’ve been through this. You can literally get whole chunks of text verbatim from these things. As shown earlier in this thread and the others like it.

Do the same with a copy machine