No one is suggesting that open source is inherently less secure
Unfortunately, I’ve met a number of people who genuinely do believe this! The same demographic who don’t know how copy and paste works or take photos of stuff on their monitor instead of print-screening and tend to end up running large corporations even though they’re completely out of touch.
There is also a lot of “security by obscurity” in the corporate/fintech world - “it’s open source so everyone can see the code which makes it less secure”. The inverse is often true thanks to Linus’s Law.
This is as transparent as hell. It reminds me of a TV show where a bunch of idiots plot to murder someone so they decide that if they all pull the trigger together, none of them are “technically” the murderer. Of course, that just meant they were all culpable.
It’s only a few layers of abstraction above “we didn’t ban these books, we flipped a coin to decide whether to ban them and fate chose tails…”
Pathetic.
TL;DR The new method still requires his art.
LoRA is a way to add additional layers to a neural network that effectively allow you to fine tune it’s behaviour. Think of it like a “plugin” or a “mod”
LoRas require examples of the thing you are targeting. Lots of people in the SD community build them for particular celebrities or art styles by collecting examples of the that celebrity or whatever from online.
So in this case Greg has asked Stable to remove his artwork which they have done but some third party has created an unofficial LoRA that does use his artwork to mod the functionality back in.
In the traditional world the rights holder would presumably DMCA the plugin but the lines are much blurrier with LoRA models.
It seems to be back now. I think it runs on a small server and quite often gets hug of deathed
Wow the enshittification is at full throttle across silicon valley! Guess those investors gotta get those returns now that interest rates are spiking!
Yeah that makes sense! I totally agree! Search is becoming pretty difficult these days!
API calls are almost always private between the caller and the endpoint (think telegram bots or mobile apps). There isn’t really a technically feasible way for a crawler to somehow “infer” any kind of knowledge of how api calls are being used unless the result has some kind of publically visible side effect (E. G. The program using the api is generating a web page and uploading it somewhere crawlable). Google et Al go by how many links from other pages to the page of interest exist (inbound links) and multiply by a smattering of other things like quality of keywords, length of content etc.
That said, if you’re implying that the api changes mean that:
That is a plausible concern.
I think IPFS often unfairly gets lumped in with crypto bro shite but it seems to me like a pretty useful technology in many other contexts too.
I think it’s a fair concern. We’ve seen other parts of the fediverse successfully implement crowd sourced funding via patron and similar to keep mastodon servers running and I suspect if Lemmy remains “the place to be” admins will have reasonable success with a similar model. Lemmy is super efficient and can support 100s of users on a single box so I think if 1% of users paid like $5 a month you could probably still support 99% of users “for free”.
Seems like the lawyer thinks that AI models deliberately jumble the Disney logo rather than specific text/artifact/logo generation just being a weakness of these types of models. (He’s wrong, he’s attributing intent to something janky/buggy)