top 85 comments

sorted by: hot top controversial new old
[–] 159 points 8 months ago (15 children)
  • [–] 66 points 8 months ago (14 children)

    Yeah, but at least this post is interesting; it shows how godawful humanity as a whole is at detecting bots in the wild.

    2 out of 400 bad.

  • source
  • parent
  • hideshow 14 child comments
  • [–] 13 points 8 months ago

    Could simply be that only 2 have been fully banned by Reddit but most have tons of subreddit bans and/or shadowbans. On the other hand, Reddit is such a cesspit these days I wouldn't be too shocked if they just exist on Reddit shitposting slop

  • source
  • parent
  • [–] 5 points 8 months ago (5 children)

    Reddit isn't trying though. Social media is hooked by big business interests and governments, and there is overlap there. I can spot influence agents, mechanized trolls, supported by bots, you can bet they could better with their tools and analytics.

    As we've seen for the last ten years, social media only takes down bots/influence agencies researchers or others make impossible to ignore, and they've cut those researchers off from the information they were using to that effect. Now it's only agencies the US government aligned groups highlight that will get removed, alleged Iranian, and the like, a bit player.

    These inauthentic accounts vastly inflate their numbers, make advertising more valuable. Even as they make the sites less useful, and drive away real users, it's also assumed that users have no where else to go so why push back on governments and big business ratfucking the sites that can hurt them in myriad ways?

    Not until we build a fediverse that can get critical mass to take off will we see them fight for real people's use of their site.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 3 points 8 months ago (4 children)

    How would we defend ourselves from such a bot flood though?

    Let's say that we start to become competitive with one of these big tech companies user wise. What is stopping them from destroying the fediverse with bots by sewing dissent, hate and slop?

    Perhaps the answer to such a scenario would just be to splinter, defederate and sort out the bot issue with better user registration.

    Happy to hear your thoughts.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 3 points 8 months ago (3 children)

    There are a few options. The best of those is part of a larger reform of how instances, and general forums they interact on, could be run. Rather than moderators that just decide on violations, bans, etc., we have a clear set of rules, with a clear set of appeal processes for bans and the like. Culminating in a jury trial of members for that instance. Maybe a higher court to decide strictly on liability grounds for users that endanger legality even with acquittals.

    Beyond that, instances could have some sort of process, maybe even election of qualified users* to appoint censors, that would have tools to hunt bots and influence operations, and flags of users would be forwarded to them, and moderators. Any enforcement actions would go through that appeal process to prevent abuses of power or misapplications of rules.

    I'd say, do it like Rome did, for every elected position, and I will get to some others, not to elect one, but two. The first two highest vote getters each get elected with the same powers. It worked for 500 years for them.

    There are some other positions we could even do elections through. Now who qualifies for elections? We could have threads where votes of reasoned arguments determine it, votes from qualified people that pass captchas perhaps. It's kind of a chicken or egg problem with voting online if influence agents and chatbots and bots are voting is the problem. Agents could cycle through accounts and do captchas, and chatbots of llm's might already be able to complete the captchas, so that might not work.

    How else could we limit voting? Maybe just by making reasoned arguments for why we should be on the voting lists, and having users with their own positive voting record able to vote, as the bots and chatbots won't have much karma without being spotted by the censors and moderators and the like?

    So I got bogged down here, but to summarize, to appoint two censors, selected by the community for 1 year terms, or whatever, that can hunt and charge accounts for removal/banning, under clear sets of rules that can be appealed to jury trials of users of the instance. Secure online trials. Maybe tests for suspected accounts.

    The trickier part, making a system of real good faith users to be able to vote so influence agents and bots and llm's don't ratfuck the votes, jury trials, etc. There would be ways, we could even establish secure end to end encryption to verify real users person to person if the person agrees, just spitballing here,

    But to maybe think about some other elected positions every term to fulfill other functions of the community. To have the clear rules and appealable enforcements to a jury trial of real users.

    Because a censor that is vetted and possessing of some analytics tools, and moderators and administrators as they are able, would be able to hunt down suspected bots and influence agents and have them removed. Not all but a lot of them. Industry ones that work off of keyword for instance, say glyphosate bad and an influence agent with bots pops up in a half hour and argues endlessly if you argue back, it' not subtle. The ones pushing for Iran forever war as we speak, also many are not subtle.

    One more thing to add. To make a separate form of karma, that results from doing favors for the community, for others, that can be traded like favors and used to qualify people. We can have real world versions for some social media applications, to be traded like credits or money, and instance type versions, not based on votes neccessarily, but for doing jobs for the community, for acting successfully as censor, or moderator, or administrator, or whatever other functions.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 2 points 8 months ago* (2 children)

    Thanks for the elaborate write-up, it's good to engage in discussion about these things. An electoral system seems quite elaborate, but I would love to see it work. It would require quite active participation by the users, and I am not convinced that smaller instances would have the active user base for this. Even so, the idea is quite appealing.

    Vetting good faith users is indeed one of the difficult problems. We could make better captchas maybe. For example taking some pictures of household objects in certain orientations. You could also check the exif flags to check if they line up with known cameras or something. Just an idea.

    I like your idea of social credit/karma, but I have some questions. So imagine that an instance can hand out these social points and so can another instance. How would we equate the value of one system to another? What is stopping my instance from minting a bunch of social points and giving them to me to elevate my "trustworthiness" in your instance?

    I also had an idea for a trust system based on belief systems. So as per Decartes I know that I am (real), and I have met some people IRL that are also real, so whenever they message me, I have a high degree of trust that they are not bots. I would also feel relatively inclined to trust the friends of my friends, with that trust decaying as we move down the friend chain. You could sort of make a "trust tree" where you would be able to see how many trust steps you are removed from IRL validation. You could even weigh the scores and downgrade someones trust score in the tree if it turned out that one of their contacts turned out to be a bot or something like that.

    Thanks for sharing your thoughts. Genuine discussion and meaningful interactions make this a better place.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 8 months ago (1 child)

    I think your ideas are good. Vetting definately needs some workshopping as you say, I like the trust based karma, I think several kinds of karma could be useful, including operational karma as below, but trust karma for sure as well.

    As to the karma and jury trials and censors and elected officials, it could be unions of instances pool ingtogether for jury pools and sharing censors, which I think is a real key evolution of federated social media, with the changes coming from llm's and chatbots, the US government alone is making armies of them, worked by contractors no doubt.

    Having an internet that isn't dead, a place to talk, and a place with a fair set of rules that are enforced in a way you can be sure to get a fair hearing in if challenged, would at some point help federated social media get the exodus from silicon valley social media as the administration, and other countries all take illiberal turns and subjugate it.

    What is needed more than anything, is a new social media, federated interoperable with existing fediverses like lemmy, whereby innumerable instances can organize on a general forum(s) on issues they agree on, and work publicly and privately as they see fit to work towards common goals. Be it groups agreeing to do or not do something if a politicians does or doesn't do something. Or to find and groom political candidates. To crowdsource information on companies that cheat people, to set up stings for instance to companies that wage theft their employees, set up workers to go in and record them doing it, crowdsource investigation, and then use the group to push for consequences, all in secret as to the organization. That would be one type of karma, karma from actually doing something, for contributing to a mission.

    That karma could then be used like currency to get people to help you with your pet project for instance. But the groups could do investigations like bellingcat, for instance the Epstein victims that want justice and are denied it, we can organize non public groups to coordinate with them, with proper operational security, and we've plenty of people on here to help us do that proper, which they would get say Operational Karma for, and the group could run down leads, investigate, could commission hackers to earn operational karma in finding documents, what have you, find witnesses and get their statements, run down their leads, organize a timeline, etc. Then we could try to use that information, people able to use it that gets play get their own operational karma for it.

    Those are just two small examples, but this could be a big force, if set up decentralized where innumerable instances can pop up whack a mole if shut down, on a new general forum endlessly, with opsec for when it's needed, for when it does threaten monied interests and attract the ire of authorities. It would come in handy for the election too, to have groups to crowdsouce what we know, what is going on, a place to securely share with trusted communities private or public information we've come into the possession of, to decide what and when to share with journalists. To be able to work to get articles into publication when it's needed, social media posts, etc.

    For instance private groups set up to catch Republicans cheating at voting machine level. They got the inner workings, the hard drives of all the voting machines after 2020 from Lindell's website and operation he set up. They illegally got their hands on like all swing state voting machine software, and probably all the red states, maybe all states by now. AZ, MI, GA, WI, PA I think. To set up ways to catch them at it. There are a lot of possibilities.

    But things removed from politics too, all encompassing, crowdsourcing new and better ways of doing things, new inventions, people could get the help and resources to run down their ideas if they can entice others to join in, they can allocate credit to each other, and get a cut of any future proceeds from it, the site would get a cut, and it could be made and sold as a benefit corporation, not to maximize profit, but to provide a needed product or service at a good cost in an area the private sector hasn't met the needs of society, while providing a reasonable rate of return and equity for those that invested in it. It could work better than all profit, and fulfill the role government refuses to regulate business into providing or providing themselves.

    Then Consumers Unions. To pressure companies to make better products, make demands on manufacturers, not buy products if they don't change their behavior.

    Anyway, so there is the political angle, of public and private groups federated to cooperate on what they agree on; there is an investigative angle with operational karma; there is an investors/inventors type of union to provide businesses to fulfill a role the private sector and government refuses to provide by operating not only on the profit motive; and consumers unions.

    If done with the right system that has the trust of people, and is resistant to government and big business fuckery, nimble in reincarnation should any part get shut down, with operational security sooner than later, it could remake society. The best time to do this was 20 years ago, but the next best time is now.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 8 months ago

    Haha you did not answer my questions, but you are clearly passiinate about this vision and I like that. As I understand it you describe a sort of moral credit that has value within the community that hands it out. So I imagine that a board would mint these tokens? What would these tokens buy you?

    We will grant that most in the community will be commited to the cause so they will want to participate, but other than respect, why and what should I grant/sell you (you having some credit) for helping the cause. Couldn't I just grant my effort to the cause directly? I get the renown aspect, but we also have commemorative mission patches pins and stickers for that.

    So in short, I am not questioning the renown/trust mechanism of a moral credit system, but I am questioning the monetary function.

    Don't take this a rejection of the idea, the rebirth of the internet has to start somewhere and that might be here by visionaries.

  • source
  • parent
  • [–] 67 points 8 months ago (13 children)

    Unfortunately they're probably around fediverse as well.

  • source
  • hideshow 13 child comments
  • [–] 44 points 8 months ago (4 children)

    That's exactly what a bot would say.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 72 points 8 months ago (3 children)

    Absolutely! That is such a fantastic, creative, and thought-provoking comment! 🚀✨

  • source
  • parent
  • hideshow 3 child comments
  • [–] 10 points 8 months ago (1 child)

    That's true to a certain extent, but I think the fediverse isn't all that attractive to these types of people. Additionally I think we are way better prepared to handle mass bot bans and detection since we aren't as whorish here in the fediverse

  • source
  • parent
  • hideshow 1 child comment
  • [–] 13 points 8 months ago (2 children)

    So far there's no money to be made here. Influence and reach is also limited.

    If that changes at some point, it might be the end of the Fediverse. It's far too open to bots. A spammer can not only easily create new accounts on instances, they can run their own instances.

  • source
  • parent
  • hideshow 2 child comments
  • load more comments (1 reply)
  • load more comments (1 reply)
    [–] 58 points 8 months ago*

    Reddit has shown through its actions that it's more interested in banning real users than bots, and wants to protect bots from being identified and called out by users, so it's not that surprising they've been able to do this.

  • source
  • [–] 46 points 8 months ago* (last edited 8 months ago) (11 children)

    The days of having arguments with Internet strangers and knowing they aren't a bot are officially over. It's hard to tell exactly when the period ended, but it's definitely done now.

  • source
  • hideshow 11 child comments
  • [–] 15 points 8 months ago (5 children)

    I did that only twice and it never did it again. Arguing with people on the internet is pointless to begin with.

  • source
  • parent
  • hideshow 5 child comments
  • [–] 2 points 8 months ago

    Plus chatbots now are getting more sophisticated, the government sponsored influence operations have the newest generation chatbots too, Israel of course the main one, but other subjects as well, like hyping the US military after the venezuela operation, which was in fact a military coup, the US made a deal with their generals to stand down and give up a few loyalist units and the presidente in exchange for the military being the de facto ruler, civilian government puppets now. Helicopters are vulnerable to weapons systems the army couldn't take our reliably without an agreement to stand down, not the least in one of the most protected places in the country when mobilized.

    But after that operation, when they were pretending the US did it under the nose of the military and not in cooperation with them, as they were threatening Cuba and Colombia, and Canada and Denmark, and Iran, and Yemen, et al, their chatbots and influence agents were running wild, entire divisions of mechanized trolls and chatbots. It was not subtle. Argue with them too much they will mass flag you and reddit will side with them, not you.

    Despite us being the ones that drive use, and the bots and agents making a less enjoyable, less useful, and less used site. They don't care, they can't see past the next set of financial statements. Bots inflate their numbers, and the government and powerful interests can jam them up if they don't lick their boots and pretend to believe them, ban the users they tell them to, etc.

    They figure we have no where else to go, so they don't have to cater to us anyway. Let's make them wrong about that.

  • source
  • parent
  • load more comments (1 reply)
    [–] 46 points 8 months ago (5 children)

    And yet I get constantly shadowbanned there just for using a VPN...

    I think reddit likes bots more than it likes real users.

  • source
  • hideshow 5 child comments
  • [–] 3 points 8 months ago (1 child)

    Bots nowadays use residential proxy networks. When people use a free VPN or other shady software, they might become part of the network, and bad actors can route traffic trough their devices.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 8 months ago*

    It's quite hard to buy residential proxies though. Almost every company selling them cheats and lies on the product, and the ips they have are absolute garbage as many people ruined their reputation already

  • source
  • parent
  • [–] 88 points 8 months ago (3 children)

    Such an inefficient way to astroturf. Just copy old comments and markov-chain basic shit. Reddit has been mostly bots for years and years.

  • source
  • hideshow 3 child comments
  • [–] 26 points 8 months ago (1 child)

    Too efficient. Why not rage bait AND raise your neighborhoods energy prices at the same time?

  • source
  • parent
  • hideshow 1 child comment
  • [–] 12 points 8 months ago

    Too hard. Have an LLM summarize each comment in an old comment chain so that it obliterates any meaning and burries any real engagement. (I have no evidence, but I think Reddit is scraping external sites and turning posts into comment chains)

  • source
  • parent
  • [–] 38 points 8 months ago

    Everyone is cooked, you are all cooked

    Thanks for making the problem worse, fuck you too man.

  • source
  • [–] 21 points 8 months ago*

    This guy just openly admitted to shitting in the global punch bowl.

    It would really be a shame if someone everyone sent an army of bots to antagonize him at every waking moment of the day.

  • source
  • [–] 19 points 8 months ago (1 child)
  • [–] 16 points 8 months ago (2 children)

    according to his own claim, and he's selling his super secret methods. he might be just making shit up

  • source
  • hideshow 2 child comments
  • [–] 15 points 8 months ago (1 child)

    Reddit is not dying fast enough

  • source
  • hideshow 1 child comment
  • [–] 14 points 8 months ago

    Godd damn clankers

  • source
  • [–] 7 points 8 months ago

    Luckily it seems the humans still feel a need to divulge their antisocial behavior.

  • source
  • [–] 6 points 8 months ago

    Little late

  • source
  • [–] 6 points 8 months ago

    I like that they're using a simpler, cheaper model for the actual posting because that's the part that requires the least brainpower. Says a lot about social media.

  • source
  • [+] 5 points 8 months ago (11 children)
  • [–] 18 points 8 months ago* (10 children)

    Clawdbot is an AI that takes full control of the PC, can open browsers, read pages, send an email, delete files, operate the CLI, install programs, anything you can do on a PC they can do. A farm is a group of PC's/servers.

    So this is a group of AI ran computers, being use for content manipulation on reddit.

  • source
  • parent
  • hideshow 10 child comments
  • [+] 4 points 8 months ago (9 children)
  • [–] 18 points 8 months ago (6 children)

    Money, Reddit makes money by selling advertising.

    They don't care if it's quality or not, more comments/posts equal more money. Capitalism 101.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 4 points 8 months ago (2 children)

    Reddit's advertising clients would see that the clickthrough rate is shit. Which dictates the price that they are willing to pay for the advertisement.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 8 months ago* (1 child)

    Have you ever clicked on a reddit ad? Many users have shit stats, and if that was used for detecting bots, then bots would click ads, reddit would lose trust, and that would damage them even more

  • source
  • parent
  • hideshow 1 child comment
  • [–] 3 points 8 months ago

    They don't care to a certain extent. But there is a threshold that if you get too many bots, companies using your platform to advertise will notice a fall in sales. And if sales drop on your platform, the money stops. Because bots don't buy things, real humans do.

    What is the threshold? I don't know, I didn't stay at a Holiday Inn last night.

  • source
  • parent
  • [–] 4 points 8 months ago

    each one of them running on ARM chips in the cloud

    "We don't want to deal with IP rate limits the hard way"

  • source
  • [–] 4 points 8 months ago (2 children)

    What does this asshole do with these bots, run influence operations? For whom? What do we know about which influence operations are hired for which interests? Forget the government ones for a moment, what about commercial interests?

    Toxic chemicals for instance, if you mention one it flags it and sends it to agents that cycle through fake accounts, backed up by bots to vote with them, to argue endlessly.

    It's like mentioning voldemort. Try it, talk trash about aspartame on reddit, or roundup, or atrazine, or god forbid nuclear energy, they've lots of real world dupes on the latter there, the decades long influence operations have borne fruit, but they have influence agents for all of those things, on keyword, the worst are on fluff pieces, propaganda pieces their PR firms or whatever make and they post it, those will be overrun, and the agents will mass flag you if you strongly argue against their bullshit.

  • source
  • hideshow 2 child comments
  • [–] 3 points 8 months ago

    What does this asshole do with these bots, run influence operations? For whom? What do we know about which influence operations are hired for which interests? Forget the government ones for a moment, what about commercial interests?

    If it's the same guy that was posted about a while ago, they do it for (blackhat) marketing. The bots mostly post comments in threads that look like a normal discussion, but where the goal is to move people away from one solution to another one. Imagine if somebody asks whether a piece of software is good, a bot then replies that they have not heard good things about it, and another bot chimes in and says "yeah I have been using instead".

    They'll probably also make manual posts claiming how good something is, put it in their control panel and a single bot will post it, while the other bots chime in with upvotes and discussion.

    Add a bit of logic to chime in on unrelated posts to make the account look more legit and you got yourself an army.

  • source
  • parent
  • [–] 2 points 8 months ago*

    Can "AI" slop really be so much worse than endless capitalist/state propaganda? Probably an improvement.

  • source
  • [–] 2 points 8 months ago

    This gives me so many ideas

  • source
  • [–] 1 point 8 months ago

    Bro I had AWS reddit bots in like 2012 you aren't clever.

  • source
  • [–] -2 points 8 months ago (2 children)

    it maybe 0.5% now but reddit will catch up and ban all the bots eventually, they find bots pretty easily. but the ones that they do not go hard ban on are propaganda ones.

  • source
  • hideshow 2 child comments