The kicker is that in any member state that’d be politics trying to bury something in procedure, while on the European level it’s the only way to get anything done. Leave out a step and the thing just fizzles because noone even knows about it.
…not, I mean, quitting X. But in general.
lemm.ee currently even proxies links to external images, sunaurus identified some issues with it, storage wasn’t one of them. Storage requirements are going to look quite differently if you’re lemmynsfw.com but it’s not particularly hard to get enough donations to afford a couple extra TB a month. Much, much cheaper than paying admins an actual wage where I think the actual scaling pain will be.
hink about how angry people get at the idea of tipping for ANYTHING and then wonder how many of those are throwing significant cash at your favorite lemmy or mastodon instance per month.
a) it’s not significant amounts, it’s quite cheap, per user, to run lemmy, e.g. lemm.ee is one of the bigger instances and costs 200 Euro a month, b) tipping 20% on a bill that doesn’t even include any service is not the same as donating to a service you like. This is more like a patreon which doesn’t lock anything for non-donating users.
Yes it’s a good thing and it’s more locally-running stuff that they’re investigating. Things like fuzzy search on your history, tl;dr bot, etc.
Malware site detection would be another idea, though they of course already have a non-local solution for that. Maybe, we do have to come full circle after all don’t we, a model that can give you an estimation of how likely it is that the page you’re looking at is AI slop.
No, it isn’t. It’s integrated into the browser, and running locally.
I’m just saying that if you a) want translation and b) privacy then you want c) AI in firefox. Because, you know, translation models are AI tech, figures that natural language is too fuzzy to do in other ways.
Not just building it’s shipping by default. That is, language detection and code that displays a popup asking you whether you want to download the actual translation model is shipping by default. About twelve megs per model, so 24 for a language pair.
The result of the whole thing was project quantum. Firefox includes lots of Rust code. Servo was never intended to be a product, it always was a research platform.
You’re free to send your data to google or deepl instead of using Firefox’s included AI translate. You know, privacy, no AI in the browser, choose one.
Chances are that any new large commercial platform will enshittify, sooner or later prompting another exodus, and each exodus will at least have some people choosing a community platform.
At least there’s only a single way to tell the computer “ok, execute this command”. And you see the command written in plain text before you.
And, no, no useful interface is intuitive because computers just have too many functions. There’s no intuitive appliance in the world with more than a temperature knob and a timer knob. Knowledge is always required, be that cultural or by RTFM.
EU fines are working. Not in the sense that they would prevent companies from trying to do shit, but in the sense that they shape up once it has been levied: Understand that those 800m are a shot before the bow. If the behaviour continues, there’s going to be daily punitive fines that very quickly become very unaffordable.
I mean, what is the money being used for?
Goes towards the EU budget, reducing the amount the member states have to pay in. In other words Berlaymont doesn’t gain anything from levying fines, their budget stays the same.
It’s still quite obscure to actually mess with AI art instead of just throwing prompts at it, resulting in slop of varying quality levels. And I don’t mean controlnet, but github repos with comfyui plugins with little explanation but a link to a paper, or “this is absolutely mathematically unsound but fun to mess with”. Messing with stuff other than conditioning or mere model selection.
Trump wants adulation, not conquest. Push come to shove you can get him out of the oval office by making him figurehead Emperor, as long as it comes with immunity he’ll accept.
On a scale of Mussolini to Hitler, he’s like 250% towards Mussolini.
The problem is: Data is code, and code is data. An algorithm to compute prime numbers is equivalent to a list of prime numbers, (also, not relevant to this discussion, homoiconicity and interpretation). Yet we still want to make a distinction.
Is a PAQ-compressed copy of the Hitchhiker’s guide code? Technically, yes, practically, no, because the code is just a fancy representation of data (PAQ is basically an exercise in finding algorithms that produce particular data to save space). Is a sorting algorithm code? Most definitely, it can’t even spit out data without being given an equally-sized amount of data. On that scale, from code to code representing data, AI models are at least 3/4th towards code representing data.
As such I’d say that AI models are data in the same sense that holograms (these ones) are photographs. Do they represent a particular image? No, but they represent a related, indexable, set of images. What they definitely aren’t is rendering pipelines. Or, and that’s a whole another possible line of argument: Requiring Turing-complete interpretation.
The company is based in an EU, EEA, EFTA, or DCFTA member country.
So Moldova, Ukraine and Georgia included, but not Turkey (only tariff union), neither are Belarus, Russia, UK, and much of the Balkans. Iceland is included, Greenland isn’t, Faroer should be via Iceland, same goes for Monaco via France, San Marino via Italy etc. Switzerland in particular is included because EFTA.
Not to mention ARM chips which by and large were/are more efficient on the same node than x86 because of their design: ARM chip designers have been doing that efficiency thing since forever, owing to the mobile platform, while desktop designers only got into the game quite late. There’s also some wibbles like ARM insn decoding being inherently simpler but big picture that’s negligible.
Intel just really, really has a talent for not seeing the writing on the wall while AMD made a habit out of it out of sheer necessity to even survive. Bulldozer nearly killed them (and the idea itself wasn’t even bad, it just didn’t work out) while Intel is tanking hit after hit after hit.
See there’s the stuff that happened, there’s the version that tankies want to believe (complete denial), which is actually different from the official CCP stance (“necessary and proportionate police action to ensure stability”, with the implication “enough questions, comrade, nothing more to see”), which is different from western public… myth, I have to say. Back when the stuff went down western journalists didn’t know what was happening, there were confusing reports, there were reports of violence, and then there was the tank man – taken the day after (IIRC, but definitely later and no he didn’t get run over). The collective imagination somehow constructed an image of the Chinese army rolling over students. Which is… metaphorically true, but not literally. And then the CCP is using that western imagination to spin their own tale of how the evil west is slandering them.
Lore books eh you’re giving me ideas. Hard to justify spending budget on that kind of stuff even if you have money to work with… how would one even get one’s hands on a woodprint artist? You know, the chisel and printing press kind? Imitating it is going to be hard indeed and figuring out how to do it not worth for a couple of one-off images you could just as well do without so either generating from prompt or telling the model to re-paint an input image in that style seems like the obvious solution.
I think a similar rule applies as when it comes to code, and NIH syndrome syndrome: Whatever it is that is your primary focus you should write yourself, use libraries for the rest. If you write a shooter, you’re going to write the gunplay, but can take the renderer off the shelf. I you’re writing a walking simulator that happens to have a gun somewhere but is generally focussed on graphical atmosphere, go grab the gunplay off the shelf but write the renderer yourself.
So unless the focus of your game is rummaging through books in an ancient library, go use that model.
If your stance is to have no or minimal defederation then lemmy.ml is not the right choice, lemm.ee is: Compare blocked instances here with the same here. Besides lemmy.ml apparently not cleaning up their blocks (e.g. exploding-heads is defunct) they block e.g. lemmynsfw.com. Not really political that’s just one that I recognised.
Also I couldn’t find a federation policy for lemmy.ml, while lemm.ee has clear rules.
Possibly not on its own but that’s not really the issue: Once you have a classifier you can use its judgements to train a generator. PhotoDNA faces the same issue that’s the reason why it’s not available to the general public.