So, there’s this multiplayer zombie FPS for the blind called Swamp, and the developer recently (as in the past few months) added an AI to help with the massive work of banning troublemakers who use predictable methods to subvert bans. Naturally, a lot of people distrust the AI (which became known as Swampnet), and it makes a convenient scapegoat for all the people demanding to be unbanned (when it turns out that they did indeed violate the user agreement).
In the past 24 hours, several high-status, obviously innocent players started getting banned. I predicted that someone was using their passwords, while everyone else went on about how Swampnet is clearly unreliable. I was tempted to throw around terms like dictionary attack, but decided against making such a specific prediction, especially without fully understanding dictionary attacks myself.
The developer confirmed that someone had been grabbing people’s passwords to link them to his (banned) account, which Swampnet uses to treat them as the same person. He also confirmed that the number of tries involved meant the villain was not brute-forcing it, but also that he hadn’t hacked the server or intercepted data packets, making him wonder if there isn’t some obvious list of passwords being shared or something.
Meta: I probably shouldn’t feel as good about outpredicting everyone and wisely avoiding getting too specific as I do. If I’d outpredicted the majority of, say, LWers, then it would feel way more justified, but that community’s selection pressures are not directed toward prediction power.
Addendum: I reread the discussion, and I treated the first one as a possible bug, but after the second clearly innocent banning, I decided it must be a hacker, and even jumped online a few times to see if I’d been hit. Posting only because I was afraid I’d used the word immediately in referring to my prediction (I did, and edited it out accordingly), when it took two datapoints for me to update to the successful prediction.
So, there’s this multiplayer zombie FPS for the blind called Swamp, and the developer recently (as in the past few months) added an AI to help with the massive work of banning troublemakers who use predictable methods to subvert bans. Naturally, a lot of people distrust the AI (which became known as Swampnet), and it makes a convenient scapegoat for all the people demanding to be unbanned (when it turns out that they did indeed violate the user agreement).
In the past 24 hours, several high-status, obviously innocent players started getting banned. I predicted that someone was using their passwords, while everyone else went on about how Swampnet is clearly unreliable. I was tempted to throw around terms like dictionary attack, but decided against making such a specific prediction, especially without fully understanding dictionary attacks myself.
The developer confirmed that someone had been grabbing people’s passwords to link them to his (banned) account, which Swampnet uses to treat them as the same person. He also confirmed that the number of tries involved meant the villain was not brute-forcing it, but also that he hadn’t hacked the server or intercepted data packets, making him wonder if there isn’t some obvious list of passwords being shared or something.
Meta: I probably shouldn’t feel as good about outpredicting everyone and wisely avoiding getting too specific as I do. If I’d outpredicted the majority of, say, LWers, then it would feel way more justified, but that community’s selection pressures are not directed toward prediction power.
Addendum: I reread the discussion, and I treated the first one as a possible bug, but after the second clearly innocent banning, I decided it must be a hacker, and even jumped online a few times to see if I’d been hit. Posting only because I was afraid I’d used the word immediately in referring to my prediction (I did, and edited it out accordingly), when it took two datapoints for me to update to the successful prediction.