3 ms·
Orkut, however, is a rare misfire for the survival model: because Orkut is so old and almost everything from that bygone era of Google has survived indefinitely
by gwern 12y ago
Orkut, however, is a rare misfire for the survival model: because Orkut is so old and almost everything from that bygone era of Google has survived indefinitely, it gave Orkut a 95% chance of surviving to 2018. Of course, everyone knew Orkut was a dead social network walking (even before Google+ was a gleam in Page's eye), but how do you tell the model that...?
Groups is in a similar situation, at 90%. I think that's also too high, but I'm not sure what I'd personally estimate the risk at. On the one hand, it feels like Google should be loathe to kill off dozens of thousands of mailing lists and its Usenet archive and decades of history & material; on the other hand, they sure act in every way like they want Groups to die and go away.
In general, the model seems to be accurate when it predicts something is very likely to die, but not so accurate when it's predicting something's very likely to survive. It'll be interesting to followup in 2018 and see how the overall set of predictions did and what an updated model looks like.