I know how hard it is to deploy features at scale, especially at a world-class company like Twitter. But if Twitter were to retain me as a consultant, I’d happily bet you any sum of your choosing that I’d make a pretty big dent in the crypto spam within a few months of full-time effort.
For those blindly downvoting that, here's what Paul Graham said:
"Either (a) Twitter is terribly bad at detecting spam or (b) there's something about Twitter that makes detecting spam difficult or (c) they don't care.
Based on my experience detecting spam, I'd guess (c)."
"Twitter engineering: If you're going to do such a bad job of catching spam, how about at least giving us a one-click way to report a tweet as spam and block the account, like email providers do? It may even help you get better at filtering, since more reports = more signal."
They’re likely downvoting my somewhat dubious claim of being able to singlehandedly make a big dent in crypto spam tweets.
I appreciate your gesture, but I don’t mind the downvotes. Bold claims warrant skepticism. And talking about votes makes the conversation less interesting for the audience.
But you’re right to point out that the problem isn’t nearly as intractable as it seems. There are many ways to deter crypto spammers.
Think of it this way: suppose Twitter’s stock price was inversely proportional to the amount of crypto spam (without accidentally removing genuine tweets). Does anyone believe the stock price would go down?
It’s why I suspect Twitter simply hasn’t made it a priority.
I remember reading several times that the enormous number of bots on Twitter inflates their "active users" stats and therefore their stock price, which is why they aren't fighting it.
The problem is false positives. Graham's experience of combating spam involved writing a Bayesian filter for his mailbox. That's fine. Somebody misses a message and one of the two parties feels bad, but they eventually either catch up or get over it. You can't "leave" that platform.
Twitter, on the other hand, is pretty sensitive to false positives, and the vernacular is so unique that naive Bayesian filtering would destroy a lot of communities with their own vocabularies and languages. If messages start arbitrarily dropping on it, its users won't stick around.
Sure, you could absolutely knock out spam. It wouldn't be that hard. Because fighting spam isn't the hard part. It's dodging the problem of firing on innocent people that the spam is using as body shields that's the hard part.
They already get incredible volumes of criticism for what little false positives they already have. Imagine if it was normal to be put in a time-out box by a Bayesian filter that wasn't tailored for your community!
Combating spam is something that has very few possible upsides for twitter, and a catastrophic failure case. Right now, spam mostly tends to effect larger accounts, who are going to stay on the platform anyway, because it's where the people are. What little spam small accounts see is manageable, and they won't leave because of it because it's so insignificant. If suddenly they couldn't send messages to others at random and without warning? Why would they stick around, then?
I do believe Twitter isn't doing well at fighting spam, also that it's a pretty hard problem. But where do you think people will go after leaving Twitter? Is there an option?
Does it matter where they'll go? People will always find some spot on the internet to have conversations after a given platform hits the friction threshold, and some might not even go anywhere: They might just leave.
Where people will go doesn't really matter, because there are a billion places they can, and there's not always a clear migration path. Sometimes, a social platform just dies, and its communities form a diaspora on different platforms, without any "clear" successor (like what happened to Orkut), or just stop doing the whole social media thing (many Google+ contributors no longer post online anywhere).
My question was kinda selfish. I want to move now. If there are a billion places, please tell me so I can move there now before the masses arrive and even that gets ruined :)
There's no easy answer to this. A billion different places users could go doesn't mean they have active communities, and if I mentioned where I like to hang out on the internet, those areas would probably get ruined.
Instead, I'll mention two platforms some of my friends like, to avoid taking the cost of a ruined platform myself:
sqwok.im: This one is the most microblog-like of the two, but it's also the least conventional. The quality varies; the front page only looks good every once in a while.
tildes.net: This one is invite-only, which helps mitigate the masses jumping in somewhat, and is run by a former reddit administrator. The existing community isn't great, but it's good. Friendly enough people.
I’d happily bet you any sum of your choosing that I’d make a pretty big dent in the crypto spam within a few months of full-time effort.
Could you do that while also making sure no false positives happen? Elon is making claims that he wants to make Twitter a platform where people are truly free to say what they want, so any spam that gets removed that isn't spam would be seen in a very poor light.
Eliminating spam when you're happy to have a few other things end up in the spam folder by mistake is relatively simple. Likewise, eliminating most of the spam but letting some through because it looks real is also quite easy. Eliminating only spam and nothing else is significantly harder.
The reason it's a ludicrous claim is that while there's overlapping skill sets, the idea that you can take an AI/ML engineer working on perception or prediction or planning problems and trivially apply them to textual content filtering is laughable. They could certainly pivot across, but it would take them a good chunk of time to get up to speed, you'd be better off hiring an engineer more familiar with that particular space.
The fact that they specified it being an AI Tesla engineer is super cringe Tesla-bro stuff and the fact that the commenter would say something like that hurts their credibility in making the other fairly extreme claims.
It's not just that they're wrong, it's that they're talking nonsense. An ML engineer working on perception models can't be trivially retasked to an NLP spam filtering problem.
If they're obviously talking out of their arse on one point then it definitely suggests they're talking out of their arse on the rest of it.
It's hard to take any of the rest of this seriously after that "detail".