4 ms·
Reminds me of the early days of spam fighting. It looked so easy: just build a Bayesian model and boom! Spam begone! Sadly, there's a drawback to open systems
by 1024core 9y ago
Reminds me of the early days of spam fighting. It looked so easy: just build a Bayesian model and boom! Spam begone!
Sadly, there's a drawback to open systems (like this one), that the robot controllers can keep probing for weaknesses and keep changing their style.
Of course, such extension also begs the question: how will they (the creators) make money? "Volume", you say?
- Bartweiss 9y agoIt's also reminiscent of tracking payment fraud. Algorithmic approaches are potent, but they're not a stable fix when your problem is adversarial and open-ended - scammers have nigh-unlimited time to hunt for new angles. As I remember it, that's a major piece of what got Palantir started - PayPal's fraud-fighting system tracked known attacks algorithmically, but they needed visualization tools to locate new attacks. That said, there's definitely an argument for attacking the ease of creation and longevity for bot accounts. Payment scammers are an endless problem because they profit on individual wins; bot networks need some sort of protracted presence to influence people.
- theashbhat 9y agohey! One of said creators here! We haven't figured out how to monetize / properly fund this yet. It's more so a problem we saw and attempted to solve. The cost right now is a couple hundred dollars month in server costs.
- 1024core 9y agoPlease don't take my comment(s) as criticism: you're doing a good job. Even if you don't make any money doing this, it's a fun thing to do. In the worst case Twitter will (acqui)hire you.