zlacker

[return to "Leaked OpenAI documents reveal aggressive tactics toward former employees"]
1. istjoh+Od[view] [source] 2024-05-22 23:37:15
>>apengw+(OP)
I wonder if this HN post will get torpedoed as fast as the one from yesterday[0].

0. >>40435440

◧◩
2. user_7+ei[view] [source] 2024-05-23 00:01:52
>>istjoh+Od
I really wish there was some simple calculations that could be shown on how posts are ranked. For eg post A has x upvotes, y comments, is z minutes old and therefore rank 2. Post B has these values, while C is here. Hence this post went down the front page quickly.

It's not that I don't trust the mods explicitly, it's just that showing such numbers (if they exist) would be helpful for transparency.

◧◩◪
3. casefi+Ak[view] [source] 2024-05-23 00:16:11
>>user_7+ei
People always interested and fascinated by the algorithm whenever it comes up. Dang makes the (correct) assertion that people will much more easily game it if they know the intricacies. PG always churlishly jumps in to say there’s nothing interesting about it and any discussion of it is boring.

Pretty asinine response but I work in Hollywood and each studio lot has public tours giving anyone that wants a glimpse behind the curtain. On my shows, we’ve even allowed those people to get off the studio golf cart to peek inside at our active set. Even answering questions they have about what they see which sometimes explains Hollywood trickery.

I’m sure there’s tons of young programmers that would love to see and understand how such a long-lasting great community like this one persists.

◧◩◪◨
4. heavys+oU[view] [source] 2024-05-23 05:49:24
>>casefi+Ak
> Dang makes the (correct) assertion that people will much more easily game it if they know the intricacies.

Which is interesting, because it's sacrilege to insinuate that it's being gamed at all.

◧◩◪◨⬒
5. dang+LV[view] [source] 2024-05-23 06:04:34
>>heavys+oU
It's not sacrilege, it's just that people rarely have any basis for saying this beyond just it kind of feels that way based on one or maybe two datapoints, and feeliness really doesn't count. We take real abuse seriously and I've personally put hundreds (feels like thousands) of hours into that problem over many years - but there has to be some sort of data to go on.
[go to top]