SnotFlickerman

joined 2 years ago
[–] SnotFlickerman@lemmy.blahaj.zone 16 points 9 months ago (2 children)

True, but the issue to me is ASRock blaming the entire problem on debris in the socket.

https://www.asrock.com/news/index.us.asp?iD=5612

After cleaning and removing debris from the CPU socket

[–] SnotFlickerman@lemmy.blahaj.zone 27 points 9 months ago (6 children)

Damn, and I had been starting to think ASRock was a pretty solid motherboard maker at this point... Not a good look...

[–] SnotFlickerman@lemmy.blahaj.zone 4 points 9 months ago* (last edited 9 months ago)

They may not, but the company is physically in the USA and their servers are physically in the USA... So... We'll see I guess.

[–] SnotFlickerman@lemmy.blahaj.zone 5 points 9 months ago (2 children)

Which is weird since the list is made up from other RSS feeds...

[–] SnotFlickerman@lemmy.blahaj.zone 8 points 9 months ago (2 children)

https://rufuspollock.com/papers/optimal_copyright_term.pdf

June 15, 2009

Using existing data on recordings and books we obtain a point estimate of around 15 years for optimal copyright term with a 99% confidence interval extending up to 38 years

Some of us have been waiting for copyright laws to be amended downward for 16 years now.

I'm not promoting that corporations should get a free pass, I just want them to be held to the same standards they held the Pirate Bay to if we're gonna pretend that current copyright laws are good, since the centerpiece of the court case against the Pirate Bay was that they were making money from what they did. OpenAI is making shitloads of money from what they did.

But I'm all for shortening copyright, but not getting rid of it. Reforms don't have to be pro-corporate slop.

[–] SnotFlickerman@lemmy.blahaj.zone 22 points 9 months ago (1 children)

But they don't have to rely on personal connections to rig the price of potatoes: they do it through a third-party data-broker called Potatotrac. Each cartel member sends all their commercially sensitive data – supply costs, pricing, sales figures – to Potatotrac, and then Potatotrac uses that data to give "advice" to the cartel members about "optimal pricing."

This is the real sick stuff, same with RealPage. They're just offering a service that could allow the businesses they serve to collude, but because they're just doing it through a third party service it's suddenly not collusion.

Doctorow pretty spot on as usual. I'm glad he's come a long way, because I actually kind of disliked his writing on Boing Boing in the early 2000's because he often got some simple facts wrong. He's much more thorough and rigorous now.

I understand what you're getting at, and this article was the best I could come up with. I think the real problem is that OpenAI is tight lipped about what they allow and don't allow. As I said, I don't personally use them, so I'm unfamiliar with if all restrictions are gone or if this is people doing the classic work-around-a-keyword filter. I have a friend who is exceptional about getting past their keyword filters in which he has done things he is definitely not supposed to be able to do.

I'll see if I can get a hold of him later tonight, because he was generating some stuff in a Ghibli style in the last few days. I'll ask if the keyword filter is still there and whether this is people just working around it, he would know better than I with first hand experience. Because I am having a hell of a time finding articles that actually detail what changed here.

I think we both want an answer to the same questions but the available writing on such questions is very limited, it seems.

[–] SnotFlickerman@lemmy.blahaj.zone 5 points 9 months ago* (last edited 9 months ago) (3 children)

https://bleedingcool.com/comics/chatgpt-wont-copy-artist-styles-including-jim-lee-frank-frazetta/

This suggests that all they've ever actually been doing is blocking keywords of artists names, and that it has always been trivial to get around such restrictions if you know how to prompt correctly.

I can't find anything about Ghibli or Miyazaki's names being on that restricted list.

Also if keyword blocking is the best they could muster, they were never serious about blocking certain styles.

From the article listed, a quote from ChatGPT:

Our policy restricts creating images in the style of artists, creative professionals, or studios whose latest work was created after 1912. Jim Lee's work falls well after this cutoff date, hence the inability to generate an image based on his style

[–] SnotFlickerman@lemmy.blahaj.zone 7 points 9 months ago (6 children)

For the longest time OpenAI’s systems would try to block people from generating images in the style of certain artists. This was obviously for copyright reasons, the didn’t want to get sued (even more than they already are). Which is something they just changed very explicitly. You can now easily generate stuff in the style of Studio Ghibli and Sam Altman made his avatar on X-The Nazi Network a ghiblified version of himself.

I don't have specifics if they have allowed other styles to be used now, too. I don't use this nonsense, but it's clear that Ghibli was put front and center.

[–] SnotFlickerman@lemmy.blahaj.zone 219 points 9 months ago* (last edited 9 months ago) (14 children)

So glad people finally waking up to these things being power plays.

Republicans, Evangelical Christians, and now Techbros are all running on the same script which boils down to "rules for thee, not for me."

Being a hypocrite is simply showing others you have the power to be a hypocrite and all they can do is get mad and stomp their feet. It's why the right wing loves to "trigger liberals." It's not even about actual politics or religion anymore, it's just simply "might makes right."

These are expressions of power, plain and simple. They should always be viewed as such.

I mean, so many companies pirated tons of materials to train their LLMs and they are making way more money than the guys at the Pirate Bay ever did. It's almost like because the guys at the Pirate Bay were making small potatoes money that they were worth going after. It's almost like if you crime big enough, the world will just pat you on the back and say "good job" instead.

Meta was literally caught downloading Anna's Archive and the widely used by nearly every AI company books3 corpus was everything from private torrent tracker Bibliotik. Why do they get different treatment? They are leveraging the same pirated works to make money, which was the whole argument for throwing the Pirate Bay admins behind bars for laws that didn't actually exist in their home country, that they were profiting from piracy. The LLM companies just are making way more money so it's let go for some reason.

It's a power play, to show little people can't get away with it, but if you've got millions in venture capital at your back, you can do whatever the fuck you want and people will praise you for it.

I mean, I think that's their point, that Elon isn't directly involved, and he's constantly shitcanning important people who would prevent something like this happening.

[–] SnotFlickerman@lemmy.blahaj.zone 9 points 9 months ago* (last edited 9 months ago)

Job ID 121

And said, chmod 000 came I out of my mother's LLM, and chmod 000 shall I return thither: the root gave, and the root hath taken away; blessed be the name of the root.

view more: ‹ prev next ›