How about prepaid credit cards?

Psychologically speaking I think about the situation as

• a learning process rather than a destination (when you mention “perfect” that’s a warning sign)
• a spectrum rather a binary position (even a king back centuries ago or a rich CEO or a powerful politician today has limited privacy, so it’s about moving positively over that spectrum)
• a worthwhile adventure helping to better learn about other things (e.g psychology, technology, politics) rather only costs

So… yes in fine it’s the same, i.e “more hoops” to go through to do the same things, BUT when framed positively it’s genuinely more exciting, more empowering!

that can help notice a compromised CDN, but not a compromised server.

Not sure I understand the distinction, a CDN is a server, so if OP is hosting code to execute on their server, they would be checked by whatever has already been downloaded and run locally before, i.e a PWA

If the hash is permanently stored in the browser, that is better, but there are also browser updates

I’m rather sure that localStorage persists over browser updates so that can be “permanent enough”

to say nothing of exploits.

I mean… sure but at that point the same apply to native. If you can’t trust the running environment you are screwed anyway.

attacker takes over the server and replaces the JS with a backdoored version, which the users receive next time they reload the page

Isn’t it exactly what hashing of JS libraries is for? e.g https://developer.mozilla.org/en-US/docs/Web/Security/Subresource_Integrity that one can see on e.g https://cdnjs.com/libraries/three.js giving you the script to execute, yes, but also a hash to verify that what you receive is indeed what you expect?

So, assuming there is once a trusted loaded version (which HAS to be the case anyway otherwise you can’t start, the same as one would do with a native executable) then there can’t be an arbitrary version loaded next without it being validated first.

PS: I’m not saying this what OP does, I’m saying executing code (Javascript or not) that must be downloaded first is not in itself a security problem.

Tinkered with a https://www.banana-pi.org/en/banana-pi-sbcs/175.html recently and… it’s really cool to have that at home, like, it works! In itself that’s quite a feat. Yet… to become actually usable due to “just” raw power but also to be economically comparable to mass produced other architectures from other manufacturers is indeed quite some road ahead.

Because it’s a race I wish such articles would bring forward comparison points :

• benchmark allowing to pinpoint past equivalents
• when was the first equivalent actually put on sale (and where)
• what’s the volume produced, even if only an order of magnitude
• inflation adjusted (as it might be several year gap) price comparison

otherwise it mostly feels like tech-propaganda pieces.

Still hitting their servers. So not doing much privacy wise

I wouldn’t underestimate how much they are getting, technically but also legally, from a logged-in account using their interface. So using another interface and without having an account can already help a lot. They don’t want “just” the data to improve a profile, they also need some way to server back the ads to, otherwise it costs them but doesn’t bring money back. I imagine in such cases, especially in jurisdictions where ghost profiles are illegal, this does a lot already.

Why is anyone surprised that the country […] that has historically dominated the Top500 list, has the fastest supercomputers?

Because since then bans have been issued, specifically preventing the purchase of the “best” hardware, and that said country does not produce such hardware internally (e.g NVIDIA and AMD top of the line, and upstream with ASML). That’s what why it is surprising, precisely because the situation has changed, cf e.g https://www.foreignaffairs.com/china/limits-china-chip-ban leading to possibly counter intuitive effects.

I imagine most people would like to better understand what hardware is being used, especially chips and to know where they come from, i.e

• are they still somehow top of the line the country can’t have through normal channels
• somehow an order of magnitude of older chips they can legally purchase, so wasting quite a bit of energy but still similar results
• the most unexpected using own hardware that is believed not to be available at scale

So yes it’s arguably surprising because the situation is not as it was just a couple of years ago.

I don’t get the hype around LLM, it is a terrible way to search

I’ll be playing devil’s advocate here just for a moment (despite the huge ecological, moral, political and economical costs) :

• what LLM does provide is a looser linguistic interface. That means instead of searching for exact words, one can approximately search for the “idea”. That means instead of hitting just the right keywords that an expert might know, one can describe a partial solution, a very rough guess of what the problem might be, and possibly get a realistic sounding answer. It might be wrong yet it might still be a step in the right direction.

So… yes I also don’t think the hype is justified but IMHO it’s quite clear that providing a solution that makes an interface easier to get some OK-looking result would appeal to masses. That means a LOT of people get their hopes up about potential empowerment and a few people ride that bubble making money on promises.

PS: for people interested in the topic but wanting to avoid the generative aspect I believe https://en.wikipedia.org/wiki/Semantic_search is a good starting point.

“brute force is brute force” what a strange thing to say, it precisely is NOT.

If you have a lot of processors but they are poorly linked together, i.e low bandwidth, then they are NOT more powerful. That’s why e.g NVIDIA is selling InfiniBand and other very expensive solutions to datacenter.

Sure a supercomputer might have more CPU/GPU/etc than another but it doesn’t make it automatically more powerful, in term of what can actually be computed in comparable time (and arguably energy consumption).

That being said, China might be secretly #1 on TOP500 but until evidence of it is provided, I’m not sure what’s the point of such speculation is.

EFF nicely provided a direct link to disable it https://x.com/settings/grok_settings and takes 2 clicks.

If all you need is to grab your groceries etc from the next village, then yes it looks like it could do that.

This is exactly the kind of usages I imagine the market target is. Namely I believe it’s :

• NOT for going from a city proper to another, e.g NOT to go from Rennes to Paris where a “big” car or train would do, even less going further
• NOT for going within a city, e.g Rennes, where public transport is rather well connected

but rather, as you suggest, going from one small town to another, say 50km radius or less. It’s while one lives in the country side to go to the farmer market on Thursday. It’s to go from and to work from the suburb, without proper bus, even less tram, to work downtown, etc.

I imagine it’s basically where most people who wouldn’t feel “adventurous” enough to use an electric bike, due to the bad weather or workload, could use something just a big bigger.

So you’re saying they are legal, truly sold, but the volume? weight? autonomy isn’t enough?

Sorry if you specified a criteria rather than an example that I missed. I’m genuinely curious as to understand because it seem you are dismissing it as useless for anyone rather than, like a buggy, something that one potentially useful but only within some context, to go with your example something one wouldn’t use in a city center but works perfectly on a beach.

PS: full disclosure, I don’t have that car, not have any economical link to the company, only trying to understand the position.

I’m not sure what you mean by “serious” here. Are you saying it’s fake in the sense that it won’t be sold? Or that the license plate would not actually legally allow it to on the road in France or Europe? Or some of the criteria, e.g autonomy, power, etc would make it realistically usable for any use case except literally playing in a playground?

I’d be up for a review on new models, e.g https://kilow.com/pages/la-bagnole from France, which claims to be repairable.

Wasn’t exactly security but I used a RPi Zero with a camera to monitor my 3D print. It’s small, low-power, wireless, didn’t have any problem with it. I imagine the result can be recorded, analyzed, etc.

IMHO the question isn’t as much you as a user of such platforms is “f*cked” because you sound both mindful and technically savvy. So, on that front, you will be OK.

The harder question I would say is how morally bankrupt you will feel by contributing to worsening the privacy of others for profit. Namely that yes by using Facebook/Insta/TikTok/etc you will gain more customers but those customers are gradually losing their privacy while you make those companies bigger by paying them. That means you depend on those companies more while they get more power.

Because of that I would argue that sure, do everything you can to protect yourself but it can’t stop there. I would argue then than the question is rather, where else can you find more clients, and maybe even “better” clients who are more aligned with your own views on privacy, and maybe even more. It’s definitely a challenge, especially seeing the trend of surveillance capitalism, but as you acknowledge yourself by using Lemmy, there are actual alternatives.

Well that’s one position, another is to say AI, being developed currently, is :

• not working due to hallucinations
• wasteful in terms of resources
• creates problematic behaviors in terms of privacy
• creates more inequality

and other problems and is thus in most cases (say outside of e.g numerical optimization as already done at e.g DoE, so in the “traditional” sense of AI, not the LLM craze) better be entirely ignored.

Edit : what I mean is that the argument of inevitability itself is dangerous, often abused.

FWIW I do have my own page on FLOSS AI, cf https://fabien.benetou.fr/Content/SelfHostingArtificialIntelligence so I do believe it’s at least interesting, even important, to understand what it is.

Still, AFAIK both the electricity https://www.bloomberg.com/graphics/2024-ai-data-centers-power-grids/ and even the potential for correction https://arxiv.org/abs/2404.04125 from intrinsic properties of the dataset and learning but also as its marketed https://link.springer.com/article/10.1007/s10676-024-09775-5 today make me reiterate, AI FLOSS doesn’t not automatically solve all problems of closed source or proprietary AI.

Edit: I know of Petals, I even discussed with some people working on it, and I learned about federated AI or federated learning back then, since at least 2019 (proof) so this isn’t new to me.

Interesting video and glad to see open-source suggested as a potential solution at the end… yet, it does not solve hallucinations (for LLMs), energy consumption (any form of AI) or… the fact that the hype itself is an economical and political tool at the service of a few. On the final point on regulators, I believe it’s damaging to imply that regulators are ignorant. They are not technical, indeed, but they are not supposed to. Regulators didn’t need to know how to build a plane to dictate rules that would improve safety in the industry, same for not being engineers in order to make the seatbelt mandatory. Yet, they do learn from technical experts, e.g in Europe the JRC that informs the Europeen Commission, Parliament, etc.