Skip Navigation

Posts
3
Comments
443
Joined
3 yr. ago

As He died to make men holyLet us die to make things cheap

  • That's cool!

    I'm also a big fan of what Bridgy Fed is capable of doing towards Bluesky - it does show that there is a lot one can actually do with the protocol.

    As I read the situation it's complicated. They are not inherently evil—on the contrary, I think they are trying to do good—but they are locked down by the structural chains around them. The whole thing was initiated by Jack Dorsey, and from the onset they wanted to re-create Twitter while solving what they perceived as "moderation challenges", and with the starting point that they were to create the next Twitter, not a decentralized network of services.

    Hell, wasn't the original idea that Twitter itself would become part of the network?

    When I see Bluesky today I see Twitter 15+ years ago. A lot of optimism and goodwill, but nevertheless a project that is doomed from the start.

  • There are minimum standards they'll have to abide by, but that's similar to Meta after their change of policy. It really is not enough that it should make anyone feel comfortable.

    Basically big platforms can choose between making moderation expensive, minimal, or arbitrary. Bluesky is leaning into minimal, keeping the door open for most things as long as they're legal. Reddit is leaning into arbitrary, having AI banning folks on account of upvotes. Facebook used to dabble with expensive, but have made a recent shift into minimal.

  • Fair - you could host a copy or a link (or a sort of combination between the two, I guess), but it wouldn't transfer the ownership of the original post. I'm still not sure this is such a pressing feature that I accept it as the actual raison d'etre of AT proto, especially considering how it very much exists there only in theory at best. But it is interesting technology, and something they could maybe have worked with ActivityPub to try to achieve.

    I'm glad to hear that maybe Bluesky is more decentralized than I suspect, but Bluesky engineer whose blog post you linked still links to his bluesky account on bsky.social. If running a separate instance is achievable, I would love to see people actually do it.

  • I guess that's fair, as a way to make users identifiable with the same user name all over the internet, no matter which platform they are on.

    When people sign in using bluesky on https://frontpage.fyi/, they are still bluesky accounts? Or does the account somehow transform into something that exists between both sites?

    Is there any real innovation here beyond a combination of "sign in with x service" and having your domain appear as your user name?

  • Nothing in ActivityPub says you can't move your content from one platform to another. It's just that Mastodon does not have this feature at the moment.

    Meanwhile, I'm not sure whether Bluesky has this feature or not, but it's somewhat irrelevant considering the fact that there are no other platforms to move your content to. The only thing I've actually seen from this is that you can use an URL as your username in the front-end, though it just points towards the same DID in the backend. I struggle to see what the great achievement here is.

    If this was the reasoning behind Bluesky, they could have developed a platform running on AP supporting the transfer of content between instances, and it would have been a whole lot easier than developing a whole new protocol.

    • The real user names (DIDs) are cryptic codes that are kept hidden most of the time, with your visible user name redirecting towards it. This gives the illusion that user names can be changed/transported, and that users are not locked down to one platform.
    • Content is filtered rather than censored, so that a big monopolistic actor can allow bigots on their platform but keep them out of sight of regular users. Had Bluesky been an ActivityPub hub, it could easily end up being perceived as a nazi bar. This is a benefit for Bluesky who do not want to be responsible for moderating their platform.

    They want decentralized moderation on a centralized platform. That's how on Bluesky, there's an understanding that the removal of hate speech "conflicts with Bluesky’s decentralized goals". On Mastodon, the decentralized nature is how we can show bigots the door without them getting to whine about their freedom of expression. Bluesky manages to create a problem using the very same concept by which Mastodon solves it.

    I guess this didn't really end up being a post about the benefits of AT. Oops.

  • Well, most people don't read Latin, so there's a high risk of ending up looking exactly as pretentious as the asshole one seeks to make fun of. That said, taste is individual, each for their own.

  • Or non-profits that are willing to accept money from supporters.

    The fact that we don't see this yet, and that Bluesky has accepted the amount of money they have from actors I would not want to be associated with, makes me doubt this is possible.

    Even if a non-profit wanted to operate with good intentions, the expense of running an AT proto hub would eventually prove a challenge, and the non-profit would either go under or need to start looking around for money. Meanwhile people can self-host their Mastodon instance on a Raspberry Pi.

    Regarding the alleged missing features of ActivityPub, I have tried and failed to understand exactly which feature is the AT proto folks so desperately wanted that they found it impossible to achieve through ActivityPub. The whole thing with having a mobile identity or whatever seems like a nothing burger to me - at the end of the day it just means that your user name is your DID number, and that web addresses can redirect towards that one. It's hardly some technological marvel that could never have been achieved on a less centralized protocol.

  • "The king is dead, long live the king".

    Or: Same shit, different wrapping.

  • As much as I hate to be that guy, it's worth keeping in mind that BlueSky is not really practising what they preach here. The AT protocol formally allows for a kind of decentralization, but it is prohibitively expensive to run an instance, meaning that only rich folks or those who are willing to accept money from venture capitalists will be capable of actually doing so.

    ActivityPub already existed when they started BlueSky. They chose to not make their protocol compatible. The reason is simple: They are a company, and they have a profit motive. ActivityPub is too democratic, and therefore hard to monetize. By now they have a bunch of crypto bro investors who want their money back. It's better to leave your money elsewhere.

    • "PhD level intelligence"
    • Fails at basic tasks

    I have never found an LLM model this relatable before.

  • Does the actual search engine load for anyone else? It's telling me "We're sorry, but something went wrong".

    Maybe it's the first ever Lemmy hug of death.

  • Econometrics has basically taken over for statistics in a lot of social sciences, for some reason. You rarely see a social scientist team up with a statistician - they team up with an economist, and they apply econometrics to whatever it is they are studying.

    There could be a couple of reasons. Economists might be perceived as having a better understanding of "the real world", as they are used to building predictive models around real world societal affairs, which is not really the job description of a statistician. Alternatively, it could be because they themselves are social scientists more than mathematicians, and they therefore "speak the language" of social sciences and are capable of interdisciplinary co-operation.

    I think it's a problem. More social scientists should learn to think critically of their methods and to do their own empirical research.

  • I used find it easier to be enthusiastic about these innovations if I could think of a single realistic use that isn't terrifying.

  • People say they prefer food cooked by professional chefs over fast food, yet new study suggests that’s not quite true

  • Likewise, poorly performing intelligence in a human or animal is nevertheless intelligence. A human does not lack intelligence in the same way a machine learning model does, except I guess the babies who are literally born without brains.

  • They are stating that the problem with AI is not that it is not human, it's that it's not intelligent. So if a non-human entity creates something intelligent and original, they might still be able to claim copyright for it. But LLM models are not that.

  • The archive link should work?