By the way, concerning the loose surge protector: Good thing you replaced that. If you see something like that, and it’s mains voltage, always replace it ASAP. Not because it doesn’t look nice or the RasPi gets unhooked… But because it’s a proper fire hazard. A whole house can burn down if you have a loose mains connection and that somehow leads to electrical sparks.

By the way, these Tesla P40 are super old. From 2016 or something. It’s the common trick to buy them used to get a lot of VRAM for cheap. But I think you should really know what you’re doing before investing several hundred bucks into something like this, as it comes with consequences.

I’m still a bit split on this. And whether the complexity and reliability is good enough for the use case… I mean if you don’t need N-out-of-M, but it’s just two people: cut a password or key in half. Same if it’s N-out-of-N people, you just need to make some puzzle pieces and hand them out, we don’t really need encryption and fancy maths for that. But I guess encrypting something would work, too. Just use a program or algorithm that’s likely still around when it’s going to be used. And you can always add a sheet of paper or PDF with instructions. Maybe save the executable file to to decrypt it somewhere if the solution requires software.

Last time I read something about this, they were still struggling massively with production. Yes, seems they’re able to manufacture those chips, but the yield is low, which makes these GPUs uneconomical. They were more a very expensive tech demo. But there isn’t a lot of detailed information out there. They might have made some progress since last year. I mean China isn’t stupid. And they can’t rely on Nvidia chips, so from their perspective there isn’t another way, they need to swallow that pill and invest whatever amount of money it takes. And I’m pretty sure they can do it. It’s just questionable whether they’re there yet and if they’re able to keep up. I have my doubts. But all of the actual information is mixed with propaganda and hype from all sides, so it’s hard to tell. But they’re determined and trying hard, I think that’s a fact.

But yeah, I don’t think Europe is leading in a lot of things. We should make an effort, though. Especially with the things that have become rarer. Like doing things ethically. And fostering democracy and freedom with the things we do and the technology we invent…

Heheh, yeah I agree, Europe is late to the race, but it’s far from being over. We should harness the power and lead the way. Though the path has to lead in the right direction. And I really dislike Zensursula on a personal level. Sure she likes AI to protect our security and mess with the public health sector. But boy does this come with issues and the potential to lead towards a dystopian sci-fi world. And since Frau von der Leyen also likes Frontex, likely predictive policing, total surveillance of all German citizens… I’m really not sure if the future she has in mind aligns with mine. Her speech is good, though. Innovation, progress and open-source are all good things. But she’s really good at speaking like a politician. And for example featuring $10 billion as the largest investment in AI is just correct since she adds the small restriction of talking about the public sector. This pile of money is completely dwarfed by private investment. And it’s unlikely to be still true in a few days time. Or it’s not even true as of today, since China has a whole coordinated public strategy for quite some time now. And politicians from the USA also already promised huge piles of taxpayer money to their AI companies…

Nice. Thanks. Seems I’ve missed some Harry Potter themed stuff. That gave me an idea… Take (or write) an Arduino library (or SSS implemeted in plain C, instead of Go), flash it on a microcontroller like an ESP32 and you have some actual, physical horcroxes. I’d have to think about the form factor, and whether they need displays, or act as a USB thumb drive… But they could light up once you get like 3 of them in bluetooth proximity and reveal the secret. Other than that I think it needed to be part of some well-maintained password vault app. Or be a web service, so people don’t need to worry to get some old computer code running.

Edit: Seems the Bitcoin people have had a thought at something like this: https://github.com/satoshilabs/slips/blob/master/slip-0039.md

Sure. I believe that could be done with minimal effort. Either by a smarthome solution, a script on a wifi router, a script in the autostart of the laptop someone uses every day, or like tasker on a phone. But you need to get it right. Or it’ll fire once you’re on a 14 day trip through Europe (and absent from your house and computer), phones can be lost or replaced… You might move… And you kind of want to make sure it’s robust enough so it actually works once needed, and that might be decades from now…

Well, I always dreamt about encrypting my master keys to all my digital heritage with some threshold scheme encryption like Shamir’s secret sharing. I believe there is some Linux tool available: http://point-at-infinity.org/ssss/

That way N out of M of my friends would have to gather after my passing, combine the puzzle pieces and be granted with access to my stuff.

There are easier ways, though. You can just write down a password and include it with your last will, seal it and have a notary take care of it. I’d create a seperate administrator account/password for that.

You could set up two factor authentification and give them one factor now, and have the other factor stored with your things so they can collect it after your passing. Doesn’t need to be complicated, create a password with 30 characters, split it in the middle and you have two factors.

There are online services for these kind of things. Or you can run some dead man switch yourself. I’m not sure what kinds of projects someone would use for that. Taking care of a dead man switch would be annoying for me.

Isn’t the chemistry of lithium batteries like well understood for quite some time now?

And the article doesn’t say much and reading the study costs at least $16.

When are we going to see something break with a big fireball? I’m a bit disappointed that they didn’t shoot themselves in the foot properly yet and the systems are still running. Shouldn’t there be a big outage? Database corruption? Or the weather forecast becoming unavailable for 3 days? Or a few more planes crashing?

Heheh, but reportedly, the hedge fund that owns Deepseek has 1.6 billion worth of servers. Plus they need to pay electricity, employees… and researchers and computer people often get a decent salary. That makes me think at least a few billions is the correct amount… If we want to do research on the level Deepseek does.

Idk, Lemmy also inreased it’s userbase by a factor of 30, mainly from a single event. It had like 1,500 MAU before summer 2023 and now we’re at 45k. So I’m not sure what to make of this. I kinda agree though, it’ll stabilize at a lower number than during a hype period. And Pixelfed aside, the more popular places on the Fediverse seem to be stagnating right now. I hope we’ve learned from the past and drama that happened and we don’t need to repeat the same things.

I don’t think that’ll work. Asking for consent and retrieving the robots.txt is yet another request with a similar workload. So by that logic, we can’t do anything on the internet. Since asking for consent is work and that requires consent, which requires consent… And if you’re concerned with efficiency alone, cut the additional asking and complexity by just straightforward doing the single request.

Plus, it’s not even that complex. Sending a few bytes of JSON with daily precalculated numbers is a fraction of what a single user interaction does. It’s maybe zero point something of a request. Or with a lots of more zero’s in-between if we look at what a server does each day. I mean every single refresh of the website or me opening the app loads several files, API endpoints, regularly loads hundreds of kilobytes of Javascript, images etc. There are lots of calculations and database requests involved to display several posts along with votes etc. I’d say one single pageview of me counts like the FediDB collecting stats each day for like 1000 years.

I invented these numbers. They’re wrong. But I think you get what I’m trying to say… For all practical purposes, these requests are for free and have zero cost. Plus if it’s efficiency, it’s always a good idea not to ask to ask, but outright do it and deal with it while answering. So it really can’t be computational cost or network traffic. It has to be consent.

(And in developer terms, some things don’t even add up. Computers can do billions of operations each second. Network infrastructure can handle somewhere in the ballpark of millions(?) of packets a second. And we’re talking about a few of them a day, here. I’d say this is more like someone moving grains of sand in the Sahara with their bare hands. You could do it all your life and it wouldn’t really change anything. For practical purposes, it’s meaningless on that scale.)

Hmmh, I don’t think we’ll come to an agreement here. I think marriage is a good example, since that comes with lots of implicit consent. First of all you expect to move in together after you got engaged. You do small things like expect to eat dinner together. It’s not a question anymore whether everyone cooks their own meal each day. And it extends to big things. Most people expect one party cares for the other once they’re old. And stuff like that. And yeah. Intimacy isn’t granted. There is a protocol to it. But I’m way more comfortable to make the moves on my partner, than for example place my hands on a stranger on the bus, and see if they take my invitation…

Isn’t that how it works? I mean going with your analogy… Sure, you can marry someone and never touch each other or move in together. But that’s kind of a weird one, in my opinion. Of course you should be able to do that. But it might require some more explicit agreement than going the default route. And I think that’s what happened here. Assumptions have been made, those turned out to be wrong and now people need to find a way to deal with it so everyone’s needs are met…

I just can’t relate. Doesn’t being in a relationship change things? It sure did for me. And I surely act differently around my partner, than I do around strangers. And I’m pretty sure that’s how most people handle it. And I don’t even think this is the main problem in this case.

I just think you’re making it way more simple than it is… Why not implement 20 other standards that have been around for 30 years? Why not make software perfect and without issues? Why not anticipate what other people will do with your public API endpoints in the future? Why not all have the same opinions?

There could be many reasons. They forgot, they didn’t bother, they didn’t consider themselves to be the same as a commercial Google or Yandex crawler… That’s why I keep pushing for information and refuse to give a simple answer. Could be an honest mistake. Could be honest and correct to do it and the other side is wrong, since it’s not a crawler alike Google or the AI copyright thieves… Could be done maliciously. In my opinion, it’s likely that it just hadn’t been an issue before, the situation changed and now it is. And we’re getting a solution after some pushing. Seems at least FediDB took it offline and they’re working on robots.txt support. They did not refuse to do it. So it’s fine. And I can’t comment on why it hadn’t been in place. I’m not involved with that project and the history of it’s development.

And keep in mind, Fediverse discoverability tools aren’t the same as a content stealing bot. They’re there to aid the users. And part of the platform in the broader picture. Mastodon for example isn’t very useful unless it provides a few additional tools, so you can actually find people and connect with them. So it’d be wrong to just apply the exact same standards to it like some AI training crawler or Google. There is a lot of nuance to it. And did people in 1994 anticipate our current world and provide robots.txt with the nuanced distinctions so it’s just straightforward and easy to implement? I think we agree that it’s wrong to violate the other user’s demands/wishes now that the’re well known. Other than that, I just think it’s not very clear who’s at fault here, if any.

Plus, I’d argue it isn’t even clear whether robots.txt applies to a statistics page. Or a part of a microblogging platform. Those certainly don’t crawl any content. Or it’s part of what the platform is designed to do. The term “crawler” isn’t well defined in RFC 9309. Maybe it’s debatable whether that even applies.

I guess because it’s in the specification? Or absent from it? But I’m not sure. Reading the ActivityPub specification is complicated, because you also need to read ActivityStreams and lots of other references. And I frequently miss stuff that is somehow in there.

But generally we aren’t Reddit where someone just says, no we prohibit third party use and everyone needs to use our app by our standards. The whole point of the Fediverse and ActivityPub is to interconnect. And to connect people across platforms. And it doen’t even make lots of assumptions. The developers aren’t forced to implement a Facebook clone. Or do something like Mastodon or GoToSocial does or likes. They’re relatively free to come up with new ideas and adopt things to their liking and use-cases. That’s what makes us great and diverse.

I -personally- see a public API endpoint as an invitation to use it. And that’s kind of opposed to the consent thing. But I mean, why publish something in the first place, unless it comes with consent?

But with that said… We need some consensus in some areas. There are use cases where things arent obvious from the start. I’m just sad that everyone is ao agitated and seems to just escalate. I’m not sure if they tried talking to each other nicely. I suppose it’s not a big deal to just implement the robots.txt and everyone can be happy. Without it needing some drama to get there.

True. Question here is: if you run a federated service… Is that enough to assume you consent to federation? I’d say yes. And those Mastodon crawlers and statistics pages are part of the broader ecosystem of the Fediverse. But yeah, we can disagree here. It’s now going to get solved technically.

I still wonder what these mentioned scrapers and crawlers do. And the reasoning for the people to be part of the Fediverse but at the same time not be a public part of the Fediverse in another sense… But I guess they do other things on GoToSocial than I do here on Lemmy.

I think it’s just one HTTP request to the nodeinfo API endpoint once a day or so. Can’t really be an issue regarding load on the instances.

Did someone complain? Or why stop?

As far as I know the parent company has these datacenters for $1.6 billion. Not Deepseek itself. So it’s way more complicated than that. The truth is, we don’t know. The $6M they claimed for the final training run is far from the total cost. And AFAIK it’s more a theoretical calculation in the first place, for a hypothetical $2.00 per hour GPU renting and their claim of gpu hours. But $1.6 billions isn’t correct either. And I’ve read in another article that isn’t even operating cost. But just the cost of the servers. And yeah, they also need energy, they probably pay their employees. But it’s other projects running in these datacenters as well, also for years to come.