Intralexical @ Intralexical @lemmy.world

Posts

0

Comments

125

Joined

2 yr. ago

Idk. It's probably morally grey at best, but the military is at least honest that they kill people. Air pollution apparently kills 6.7 million people annually, compared to every war put together killing just 0.1 million people annually over the last decade. But the difference is that if you work in air pollution you get to run slick ads bragging about how clean you are, whereas if work in murder-machines you're more directly confronted with the reality and the gravity of what you're doing— The latter may be scarier, but the former feels much more gross.

I agree, but by now there’s probably no reason to make people write those kind of things. It’s likely that no human oversight is needed at all. Astroturfing can now be nearly completely automated.

Probably I'm just picking semantics, but that kinda is the good reason to make humans write those kinds of things. Astroturfing is bad, so needing to pay an entire human to be able to do it imposes a cost that limits its spread and application. I guess that's also what you're saying.

But that’s a very small gain considering the massive loss of trust in the web and making it a glitchy, spammy, scammy experience.

I've been passively wondering how long it will be until I have to start adding before:2023 to get remotely useful web search results on any topic. Don't know what to try yet if I need to look up something from after that.

On the contrary, I believe our inherent ability to trust each other is one of the main pillars of civilization, and undisclosed use of LLMs heavily undermines it.

Oh yeah, definitely. I just meant that as an ironic silver lining, the damage would probably be worse if there wasn't already some level of dishonesty and deception in society, because then we'd be too pure to have any defences against LLMs.

Apparently, this has been around for ages:

May 28 2020: https://consensys.net/blog/news/everything-you-need-to-know-about-reddits-new-blockchain-based-community-points/

May 15 2020: https://old.reddit.com/r/ModSupport/comments/gk0x6g/could_the_admins_please_explain_this_community/

IA History: https://web.archive.org/web/20230000000000*/https://www.reddit.com/community-points

Discussions of proliferating it and fantasies about the resulting weird exploitation it will enable have occurred as well:

March 17 2023: https://old.reddit.com/r/CryptoCurrency/comments/11tv5vz/when_community_points_are_introduced_to_other/

July 16 2023: https://old.reddit.com/r/CryptoCurrency/comments/151fg8h/which_subreddits_need_to_be_the_next_to_get/

Apparently, they also tried to do this before, almost 9 years ago:

https://www.engadget.com/2014-12-19-reddit-notes.html

https://techcrunch.com/2014/12/19/reddit-announces-redditnotes-a-way-to-share-equity-with-readers/

https://www.theverge.com/2014/12/20/7427491/reddit-notes-announced-give-5-million-dollars-to-users

https://slate.com/business/2014/12/what-reddit-notes-are-their-history-and-future.html

https://old.reddit.com/r/blog/comments/2pt25f/announcing_reddit_notes/

https://old.reddit.com/r/TheoryOfReddit/comments/5fyaec/what_happened_to_reddit_notes/

https://www.fortune.com/2015/01/30/reddit-notes-is-not-going-to-happen/

Kinda puts things in a new light. Ruining the site has always been the plan.

It's a standard USB device. It interacts with MacOS only via the USB bus, following the standard/spec. If it craps its pants and fails catastrophically when it receives specific, valid USB commands, then it doesn't matter whether it's only MacOS or TempleOS that sends the commands that trigger that particular failure case. It's still the drive's fault, because it's failing to do its literal one and only job of safely storing your data over USB.

E.G.: If MacOS can trigger the failure now, then there's no guarantee that any future update to Windows or Linux won't also trigger the same failure.

That is exactly the type of content LLMs were designed to excel at generating.

Hm. It's also exactly the kind of disingenuousness that that humans have spent a couple million years evolving to try to detect, though.

I wonder if the LLMs are going to win this. Maybe more likely: When everyone realizes that the entire Internet is being flooded with even more bullshit, we'll just stop trusting it, and the LLMs will more or less have put themselves out of a job.

It would be funny if the propensity for humans to lie to each other meant that we were basically already inoculated from this terrifying new category of machines that we've designed to lie to us too.

PCM, ASCII, and straight RGBA bitmap encodings aren't going anywhere. By extension, derived formats like WAV, UTF-8, and word processor files and webpage HTML are mostly fine too. The formats are structurally simple enough that even if the associated file extensions were somehow to be forgotten, all you'd need to do to invent them again is hand the file to a bored nerd over the weekend.

I think you kinda got the BBC and NASA problems backwards. The BBC's had a couple of prominent incidents where digital "preservation" that was supposed to be eternal couldn't even be opened anymore after a couple of years, like their Domesday Book/Project application thingy. They've also lost a bunch of old shows, like early Dr. Who episodes, I think. NASA didn't just forget how to read the Apollo tapes; they overwrote them to reuse the tapes, as was their standard practice at the time. The original signal and tapes were very HD (or analog), but most of the videos we have today are from the TV camera that they pointed at their own TV screen last-minute when they realized they didn't have an adapter for broadcast— The equivalent of a grainy cell phone photo of a screenshot, basically.

The BBC and NASA incidents happened in an era before computers were a ubiquitous commodity product. So, everyone and their cat was basically inventing their own obscure single-implemention proprietary file formats at that time. Nowadays we have established technical standards, as well as formats that have already sorta stood the test of time based on their utility and simplicity— and millions of people who already know how to read them— so that particular vector for bitrot isn't really as much of an issue anymore.

…That said, I think I sorta missed your point. What you're really saying is that stewardship of digital records is much trickier and riskier than stewardship of physical records— and that results in stuff being lost. And that is absolutely true.

I’ve been telling people for years that the entire 21st century is at risk of being a lost century. Even personally I can’t guarantee my data will be with me 20 years from now even though I back it up. If you care about a photo or document, print it and throw it it a box. As I get older I find more of an obsession with physical media from a preservation point of view. Because I know my books and pictures will be around 50 years from now. Digital files not so much.

LOCKSS and KISS, though. Flash chips don't last forever but are pretty durable, and so are optical media as long as they're the right material. SSDs decay and HDDs fail, but for magnetic platter media even if the head or motor crashes there's always the old magnetic microscope in a pinch. USB's not going anywhere, and if you have four or five copies that you don't completely neglect and don't store in the same physical place, presumably you'll have the chance to notice and take corrective measures if any of them start failing or are at risk.

I don't actually know that an individual book or picture will still be around in 50 years; Fire, flooding, insects, acidic paper, low-quality ink maybe— Digital stuff's fragile, but so is physical stuff. Stick it in the attic, and the heat'll speed up any chemical reactions and probably make it cozier for insects; Stick it in the basement, and the condensation will get you mildew and rot. By contrast, having a flash drive accidentally survive a trip through a washer and dryer is a pretty common occurrence, and I've yet to lose a drive even with that level of negligence. Material compatibility's one of the very most basic parts of a set of very precise manufacturing techniques, tin whiskers seem pretty rare these days, the really scarily insidious stuff like hydrogen embrittlement is super improbable, and most biological forms of decay haven't adapted to eating cured epoxy and monocrystalline silicon yet.

At least I sorta know how a flash cell or hard drive platter is meant to be structured; Who knows what weird organic reactions and unstable or slowly diffusing molecules are happening in the pile of chemical pigments on a sheet of likely-acidic bleached cellulose and cheap ink or toner, and whether it will still be legible to human eyes in however many years? Plus, a printed photo or document starts fading the very instant it's created, and it gets a little worse every time you touch it with sweaty human hands or look at it while exhaling moist human breath and corrosive enzymatic saliva droplets under a white LED lamp or G-type star shooting out ionizing UV rays. Digital failures tend to be catastrophic, but at least up until the moment it fails, you can make sure that it is the exact same picture or text— And you can make many, many copies very cheaply, all of them very physically durable compared to paper, and know that they are all the exact same picture and text.

That said, I absolutely agree with your overall assessment that most of the information in the early 21st century, including most of the public Internet/WWW, most likely either will be or already is… Maybe not technically lost, per se, given how much caching and saving happens on private clients, but certainly rendered inaccessible.

Ideally I'd really love to see a return of microfiche, actually, using modern polymers and metallization. I've been meaning to look into that for a while now. At a reasonable scale for optical viewing, you could fit… much, much more content than you might expect, and do it several times over, in an entirely reasonable number of pages. Your comment actually spurred me to finally think of a practical way of printing that— for years before, I'd been trying to idly figure out a process based on photomasks and nanoparticles suspended in resin, which had always felt like a very messy and tricky idea, but I just thought of another idea– So thanks for providing some inspiration there.