7 ms·

> One AI engineer from Tesla could solve Twitter’s bot and spam problem I think you underestimate how hard the problem is or you don’t fully understand how it

> One AI engineer from Tesla could solve Twitter’s bot and spam problem I think you underestimate how hard the problem is or you don’t fully understand how it works. The scale at which Twitter operates makes it near impossible to take down spam in real time. You don’t think Twitter is working on it, but they are. What you see is a much smaller percentage of the actual spam generated, and the reason you see it is because users are creative on circumventing all the preventive actions they take.

eldenwrong 5y ago
Add a bitcoin deposit to verify account as orange. Problem solved
viraptor 5y ago
An alternative take: there's spam that's trivial to spot and spam that's hard to spot. Twitter is ignoring both. I can give you both searches and keywords to post which will return lots of spam ready to block. Not everything of course, but let's not post the "spam is a hard problem" excuse until they nuke every account responding with "metamask" and a link to Google forms.
- soperj 5y ago
  You don't nuke the accounts, you make it so no one can see their posts and they don't realize. Make them waste their time instead of creating a new account.
  - coffeeblack 5y ago
    You do neither. You offer users a list of filters they can choose to activate, so they only see content according to the filters they enabled. There would be filters against spam, filters against opinions other than your own, etc.
  - viraptor 5y ago
    Implementation detail ¯\_(ツ)_/¯ you can handle this in many ways. The point was that Twitter doesn't currently care to use even minimal effort.
    - ben_w 5y ago
      We don’t know how much effort they’re putting in. What’s the false positive rate, and is that why people complain they’re being banned or shadow banned? (Definitions of shadow ban vary: https://blog.twitter.com/en_us/topics/company/2018/Setting-the-record-straight-on-shadow-banning https://blog.twitter.com/en_us/topics/company/2018/Setting-t...)
      - viraptor 5y ago
        My thesis is: if they put in minimal effort into automation, then metamask spam would be trivial to block. Since we know of the metamask spam for months (years?), it means there was no effort. I could be wrong, but I'll stick to the most obvious explanation.
        ben_w 5y ago
        Given I had to Google "metamask" to understand your comment, I’m not sure it’s as widespread as I think you think it is.
        viraptor 5y ago
        It's been going on for months - if you're not aware of this it doesn't mean it's not widespread. If you tweet "metamask problem", you'll get a number of bots immediately responding with Google forms links for "metamask support". Yes, it's very widespread and extremely obvious. Both fake support accounts https://twitter.com/metamesksuport/with_replies https://twitter.com/metamesksuport/with_replies (I can find at least 10 more "support" accounts from a basic search) and helpful randos https://twitter.com/ij851227/status/1515302472079847425 https://twitter.com/ij851227/status/1515302472079847425 It's been tweeted by multiple security people at various Twitter engineers already and raised in lots of ways. They know and don't care. Meantime, real people lose their money - otherwise nobody would waste their time to post the scam.
  - onefuncman 5y ago
    Whoops! As a result of your action, your userbase has now hyperfixated on follower counts as a proxy for status.
ehsankia 5y ago
To think that a company that has failed to create self-driving for nearly a decade could solve "content moderation", one of the hardest problem on the internet that every single of the big companies (who btw hire the best talent) have been struggling with, is truly a shocking statement to me. If all Google and Facebook couldn't solve it, why could a Tesla engineer of all people...
- paisawalla 5y ago
  We're not talking about the hardest content problems on the internet. Look at any of Elon Musk's (or anyone in crypto's) popular tweets and you'll see a flood of incredibly obvious bots that poorly mimic his account, spamming some shit coin or advertising a BTC address. Also, you're really understating how good Tesla self-drive is. It's not full self-drive but it's a real achievement.
  - ehsankia 5y ago
    Yes, and Youtube and Facebook both have similar problems. There is survivalship bias at play here, you do see the bots that manage to get past the spam protection, but never see the ones that don't. For all you know, Twitter could be blocking 99% of the spam. Of course, as long as there's money in it (and there is), spammers will constantly be coming up with new and unexpected ways to bypass your systems. It's a neverending battle, and if you think it's trivial, you are the one understating how hard the problem is. I never claimed Tesla self-driving wasn't an achievement, but clearly they underestimated how hard that would be themselves, so I wouldn't be surprised if they'd do the same for content moderation.
    - paisawalla 5y ago
      There's a substantial qualitative difference between what gets through Youtube's and Facebook's filter, versus what gets past Twitter. They're not even practicing the current state of the art, as practiced by their peers.
    - LewisVerstappen 5y ago
zackees 5y ago
I reported child pornography on twitter. It took a while for it to be taken down. Yet people talking about hunter biden’s laptop in October 2020 had their accounts deleted immediately. Twitter is very good at content moderation when it threatens the manufactured narrative of the global elite which they serve. And then completely incompetent otherwise. I don’t buy the line that the problem is too complex because I see selective competence.
xwdv 5y ago
AI Engineers at Tesla also deal with spam, just not in the way you think. Self driving AI has to deal with input spam from all sorts of details in the real world, most of which are irrelevant to making driving decisions. So yes, I’d be reasonably certain AI Engineers at Tesla could easily solve spam on a simple web platform.
prvc 5y ago
Some scammers who appear in Musk's replies impersonating him use similar or identical avatars and screen names in order to perpetrate their fraud. Removing and preventing the creation of this type of account would not even require "AI". It hasn't happened yet because Twitter lacks the will to do it, for whatever reason.
- chippiewill 5y ago
  > Removing and preventing the creation of this type of account would not even require "AI". It wouldn't require _Machine Learning_, the kind of filtering approach you've just described would absolutely constitute AI.
  - once_inc 5y ago
    Calculating the levenshtein distance from users, the average difference between the avatars, and looking at basic interactions in reply-to-reply-to-users isn't AI. That's heuristic. And it should have been implemented ages ago.
  - prvc 5y ago
    Call it what you will, but the point is it would be easy for them to do.
murbard2 5y ago
Spam on most platform is a complicated engineering challenge. However, for some inexplicable reason, Twitter features a kind of spam that you don't see on any other platforms because it could be caught with a regex. It's almost never the case that spam issues on popular social networks can be alleviated with some easy fixes, because obviously if they could they would already have done that. Twitter is a very weird exception.
Hypocritelefty 5y ago
hwers 5y ago
The spam scale problem should be solvable with a one-shot locally evaluated NN trained to classify behavior/tweets from their massive massive ground truth dataset. If it’s evaluated locally then I don’t see how scale is an issue.
misiti3780 5y ago
If they stopped reporting and focusing on MAU, the problem isnt that hard to solve. If you want to operate a bot on twitter, it costs X/month or just make bots illegal. The problem would be more surmountable at that point

eldenwrong 5y ago

Add a bitcoin deposit to verify account as orange. Problem solved

viraptor 5y ago

An alternative take: there's spam that's trivial to spot and spam that's hard to spot. Twitter is ignoring both. I can give you both searches and keywords to post which will return lots of spam ready to block. Not everything of course, but let's not post the "spam is a hard problem" excuse until they nuke every account responding with "metamask" and a link to Google forms.

soperj 5y ago
You don't nuke the accounts, you make it so no one can see their posts and they don't realize. Make them waste their time instead of creating a new account.
- coffeeblack 5y ago
  You do neither. You offer users a list of filters they can choose to activate, so they only see content according to the filters they enabled. There would be filters against spam, filters against opinions other than your own, etc.
- viraptor 5y ago
  Implementation detail ¯\_(ツ)_/¯ you can handle this in many ways. The point was that Twitter doesn't currently care to use even minimal effort.
  - ben_w 5y ago
    We don’t know how much effort they’re putting in. What’s the false positive rate, and is that why people complain they’re being banned or shadow banned? (Definitions of shadow ban vary: https://blog.twitter.com/en_us/topics/company/2018/Setting-the-record-straight-on-shadow-banning https://blog.twitter.com/en_us/topics/company/2018/Setting-t...)
    - viraptor 5y ago
      My thesis is: if they put in minimal effort into automation, then metamask spam would be trivial to block. Since we know of the metamask spam for months (years?), it means there was no effort. I could be wrong, but I'll stick to the most obvious explanation.
      - ben_w 5y ago
        Given I had to Google "metamask" to understand your comment, I’m not sure it’s as widespread as I think you think it is.
        viraptor 5y ago
        It's been going on for months - if you're not aware of this it doesn't mean it's not widespread. If you tweet "metamask problem", you'll get a number of bots immediately responding with Google forms links for "metamask support". Yes, it's very widespread and extremely obvious. Both fake support accounts https://twitter.com/metamesksuport/with_replies https://twitter.com/metamesksuport/with_replies (I can find at least 10 more "support" accounts from a basic search) and helpful randos https://twitter.com/ij851227/status/1515302472079847425 https://twitter.com/ij851227/status/1515302472079847425 It's been tweeted by multiple security people at various Twitter engineers already and raised in lots of ways. They know and don't care. Meantime, real people lose their money - otherwise nobody would waste their time to post the scam.
- onefuncman 5y ago
  Whoops! As a result of your action, your userbase has now hyperfixated on follower counts as a proxy for status.

ehsankia 5y ago

To think that a company that has failed to create self-driving for nearly a decade could solve "content moderation", one of the hardest problem on the internet that every single of the big companies (who btw hire the best talent) have been struggling with, is truly a shocking statement to me. If all Google and Facebook couldn't solve it, why could a Tesla engineer of all people...

paisawalla 5y ago
We're not talking about the hardest content problems on the internet. Look at any of Elon Musk's (or anyone in crypto's) popular tweets and you'll see a flood of incredibly obvious bots that poorly mimic his account, spamming some shit coin or advertising a BTC address. Also, you're really understating how good Tesla self-drive is. It's not full self-drive but it's a real achievement.
- ehsankia 5y ago
  Yes, and Youtube and Facebook both have similar problems. There is survivalship bias at play here, you do see the bots that manage to get past the spam protection, but never see the ones that don't. For all you know, Twitter could be blocking 99% of the spam. Of course, as long as there's money in it (and there is), spammers will constantly be coming up with new and unexpected ways to bypass your systems. It's a neverending battle, and if you think it's trivial, you are the one understating how hard the problem is. I never claimed Tesla self-driving wasn't an achievement, but clearly they underestimated how hard that would be themselves, so I wouldn't be surprised if they'd do the same for content moderation.
  - paisawalla 5y ago
    There's a substantial qualitative difference between what gets through Youtube's and Facebook's filter, versus what gets past Twitter. They're not even practicing the current state of the art, as practiced by their peers.
  - LewisVerstappen 5y ago

zackees 5y ago

I reported child pornography on twitter. It took a while for it to be taken down. Yet people talking about hunter biden’s laptop in October 2020 had their accounts deleted immediately. Twitter is very good at content moderation when it threatens the manufactured narrative of the global elite which they serve. And then completely incompetent otherwise. I don’t buy the line that the problem is too complex because I see selective competence.

xwdv 5y ago

AI Engineers at Tesla also deal with spam, just not in the way you think. Self driving AI has to deal with input spam from all sorts of details in the real world, most of which are irrelevant to making driving decisions. So yes, I’d be reasonably certain AI Engineers at Tesla could easily solve spam on a simple web platform.

prvc 5y ago

Some scammers who appear in Musk's replies impersonating him use similar or identical avatars and screen names in order to perpetrate their fraud. Removing and preventing the creation of this type of account would not even require "AI". It hasn't happened yet because Twitter lacks the will to do it, for whatever reason.

chippiewill 5y ago
> Removing and preventing the creation of this type of account would not even require "AI". It wouldn't require _Machine Learning_, the kind of filtering approach you've just described would absolutely constitute AI.
- once_inc 5y ago
  Calculating the levenshtein distance from users, the average difference between the avatars, and looking at basic interactions in reply-to-reply-to-users isn't AI. That's heuristic. And it should have been implemented ages ago.
- prvc 5y ago
  Call it what you will, but the point is it would be easy for them to do.

murbard2 5y ago

Spam on most platform is a complicated engineering challenge. However, for some inexplicable reason, Twitter features a kind of spam that you don't see on any other platforms because it could be caught with a regex. It's almost never the case that spam issues on popular social networks can be alleviated with some easy fixes, because obviously if they could they would already have done that. Twitter is a very weird exception.

Hypocritelefty 5y ago

hwers 5y ago

The spam scale problem should be solvable with a one-shot locally evaluated NN trained to classify behavior/tweets from their massive massive ground truth dataset. If it’s evaluated locally then I don’t see how scale is an issue.

misiti3780 5y ago

If they stopped reporting and focusing on MAU, the problem isnt that hard to solve. If you want to operate a bot on twitter, it costs X/month or just make bots illegal. The problem would be more surmountable at that point