4 ms·

I've long suspected there are vast bot networks like this all over social media. YouTube comment sections on videos related to covid are an absolute cess pit o

I've long suspected there are vast bot networks like this all over social media. YouTube comment sections on videos related to covid are an absolute cess pit of covid deniers, misinformation spreading and conspiracy theory peddling. I would like to think these are bots and not a very vocal but significant section of society but given recent world events I am not sure at all.

jjcon 6y ago
> given recent world events I am not sure at all. Have recent events involved significant sections of society?
BelenusMordred 6y ago
You don't need some exclusive GPT-3 access for this. A simple Markov bot is enough to fool nearly everyone they perform fairly well when given specific domains, quite poorly in general online forums though. Give that bot a foreign name and virtually every native English speaking person will assume it's simply bad grammar.
- notahacker 6y ago
  It's Twitter, you don't even need to generate original text: likes, comment-free retweets, reaction gifs and posting the same text as replies to everything are all actions that will boost whatever sentiment or link it is you want to boost.
simias 6y ago
I genuinely wonder what's the real number. It is easy to find accounts that look like they could be bots, but I wonder what portions of those are just very dedicated individuals. I don't use social media a lot but I do regularly post on HN for instance. I don't know what my average message rate is but it's probably around 1 message/day. Now if I just wanted to spam my political agenda on the site (and the site let me do it) I could very easily copy/paste the same bullshit a hundred times a day on every story on the frontpage. If like 20 other people did the same thing the amount of noise we'd generate is pretty large Look at random fan wikis to see the amount of free work some dedicated individuals can achieve. I'm certain that many of these super active troll accounts are just that, dedicated individuals who feel very strongly about something and just spend 10 hours every day spamming and harassing and copy-pasting. If your output is like 100x that of a normal user (and you know how to game the algorithms to amplify your message) you don't need a huge number of accounts or automation to become a real nuisance.
- lmeyerov 6y ago
  We research this stuff - a lot of nation-state and financed activity is farms of low-paid people assisted by software running multiple and shared accounts. Misinfo analysts (including ML tools) can suspect this bc the accounts clock in & clock out and follow a talking points script. Except bc a low paid person is adding a human touch, account abuse teams see it and do nothing due to uncertainty, despite being against TOS. Sometimes it is more obvious. People had MAGA campaign apps that lets Trump's team remotely activate otherwise-legit social media accounts. So the account is real, but popular messages were bot behavior. Twitter is even less likely to want to take down such an account, despite the prohibited behavior.
  - Balgair 6y ago
    I'm reminded of the quote: "There is considerable overlap between the intelligence of the smartest bears and the dumbest tourists" - Ascribed to an unnamed Yosemite Park Ranger. The quote may never really have happened, but I think it fits here. If the people are intentionally acting as bots per the direction of other people, then what does that make the people acting as bots?
therockspush 6y ago
I used to run a twitter bot. It didn't create any content. Just liked, retweeted or followed based on keywords. Then my tweets would get liked and followed. 95% of that was just bots. Every now and then I would unfollow a real person and my account, which was clearly a bot, would get some hate mail telling me to re-follow them. It was pretty surreal.
heidar 6y ago
100%. I find this incredibly interesting and I'd love to understand more about what topics they post about. I bet many hot topics, especially ones about politics, have a fair share of bots commenting. I think their purpose is equally interesting. I'd guess a lot of them are there to polarise discussions and drive people apart, but there are probably ones that have an agenda to drive a specific opinion as well. I think this isn't just on Twitter, but other platforms as well.
imiric 6y ago
Not just YouTube comments, but videos as well have become automated or semi-automated to produce the most engaging content and maximize ad revenue. There are companies making millions by gaming the system. See the Elsagate scandal and TheSoul Publishing[1], the company behind those 5-minute craft videos and occasional Russian propaganda. And of course similar things are happening on all social media platforms. It's deeply ironic that the West built all these tools with the intention of "connecting" people, something that's still Facebook's main mission statement, yet its biggest effects have been exactly the opposite, causing division and distrust between people, and contributing to real world conflicts and toppling of governments. Social media today is weaponized by political agents to spread misinformation and propaganda on an unprecedented scale and the platforms are paying them to do it. The current political climate in the US is no surprise. It was a coordinated psyops attack and its enemies are laughing at how easy and inexpensive it was to destabilize a nation compared to traditional warfare. I fear that we still haven't seen the worst of it, and frankly I'm not sure what the solution is. [1]: https://www.lawfareblog.com/biggest-social-media-operation-youve-never-heard-run-out-cyprus-russians https://www.lawfareblog.com/biggest-social-media-operation-y...

jjcon 6y ago

> given recent world events I am not sure at all. Have recent events involved significant sections of society?

BelenusMordred 6y ago

You don't need some exclusive GPT-3 access for this. A simple Markov bot is enough to fool nearly everyone they perform fairly well when given specific domains, quite poorly in general online forums though. Give that bot a foreign name and virtually every native English speaking person will assume it's simply bad grammar.

notahacker 6y ago
It's Twitter, you don't even need to generate original text: likes, comment-free retweets, reaction gifs and posting the same text as replies to everything are all actions that will boost whatever sentiment or link it is you want to boost.

simias 6y ago

I genuinely wonder what's the real number. It is easy to find accounts that look like they could be bots, but I wonder what portions of those are just very dedicated individuals. I don't use social media a lot but I do regularly post on HN for instance. I don't know what my average message rate is but it's probably around 1 message/day. Now if I just wanted to spam my political agenda on the site (and the site let me do it) I could very easily copy/paste the same bullshit a hundred times a day on every story on the frontpage. If like 20 other people did the same thing the amount of noise we'd generate is pretty large Look at random fan wikis to see the amount of free work some dedicated individuals can achieve. I'm certain that many of these super active troll accounts are just that, dedicated individuals who feel very strongly about something and just spend 10 hours every day spamming and harassing and copy-pasting. If your output is like 100x that of a normal user (and you know how to game the algorithms to amplify your message) you don't need a huge number of accounts or automation to become a real nuisance.

lmeyerov 6y ago
We research this stuff - a lot of nation-state and financed activity is farms of low-paid people assisted by software running multiple and shared accounts. Misinfo analysts (including ML tools) can suspect this bc the accounts clock in & clock out and follow a talking points script. Except bc a low paid person is adding a human touch, account abuse teams see it and do nothing due to uncertainty, despite being against TOS. Sometimes it is more obvious. People had MAGA campaign apps that lets Trump's team remotely activate otherwise-legit social media accounts. So the account is real, but popular messages were bot behavior. Twitter is even less likely to want to take down such an account, despite the prohibited behavior.
- Balgair 6y ago
  I'm reminded of the quote: "There is considerable overlap between the intelligence of the smartest bears and the dumbest tourists" - Ascribed to an unnamed Yosemite Park Ranger. The quote may never really have happened, but I think it fits here. If the people are intentionally acting as bots per the direction of other people, then what does that make the people acting as bots?

therockspush 6y ago

I used to run a twitter bot. It didn't create any content. Just liked, retweeted or followed based on keywords. Then my tweets would get liked and followed. 95% of that was just bots. Every now and then I would unfollow a real person and my account, which was clearly a bot, would get some hate mail telling me to re-follow them. It was pretty surreal.

heidar 6y ago

100%. I find this incredibly interesting and I'd love to understand more about what topics they post about. I bet many hot topics, especially ones about politics, have a fair share of bots commenting. I think their purpose is equally interesting. I'd guess a lot of them are there to polarise discussions and drive people apart, but there are probably ones that have an agenda to drive a specific opinion as well. I think this isn't just on Twitter, but other platforms as well.

imiric 6y ago

Not just YouTube comments, but videos as well have become automated or semi-automated to produce the most engaging content and maximize ad revenue. There are companies making millions by gaming the system. See the Elsagate scandal and TheSoul Publishing[1], the company behind those 5-minute craft videos and occasional Russian propaganda. And of course similar things are happening on all social media platforms. It's deeply ironic that the West built all these tools with the intention of "connecting" people, something that's still Facebook's main mission statement, yet its biggest effects have been exactly the opposite, causing division and distrust between people, and contributing to real world conflicts and toppling of governments. Social media today is weaponized by political agents to spread misinformation and propaganda on an unprecedented scale and the platforms are paying them to do it. The current political climate in the US is no surprise. It was a coordinated psyops attack and its enemies are laughing at how easy and inexpensive it was to destabilize a nation compared to traditional warfare. I fear that we still haven't seen the worst of it, and frankly I'm not sure what the solution is. [1]: https://www.lawfareblog.com/biggest-social-media-operation-youve-never-heard-run-out-cyprus-russians https://www.lawfareblog.com/biggest-social-media-operation-y...