3 ms·

Reading through the bots comments, and the HN replies is interesting and depressing. Comments that have substance will contradict opinions and groupthink and w

Reading through the bots comments, and the HN replies is interesting and depressing. Comments that have substance will contradict opinions and groupthink and will get downvoted. Comments that say nothing like this bot randomly walk up. But the bot pre-dates the GPT-3 official release of June. So I assume it's internally from someone in OpenAI or not GPT-3..... We shall see I guess.

aaron695 6y ago
The one on HN has been running for many months (The one on Reddit only lasted weeks) and has only been called a GPT-3 once I can see. Given it's ability to get a lot of upvotes and no need for a voting ring it would allow a lot of control over discussions through up votes and down votes. Mmmm Russia. It'd be interesting to see it's upvote and downvote history. There are definite attacks on HN you can see play out. I think this bot is actually academic since it is very prolific, it could hide itself a lot more if it wanted to. [edit] OT: I'll give an example of a ongoing Russian attack, Bellingcat is constantly flagged. https://news.ycombinator.com/from?site=bellingcat.com https://news.ycombinator.com/from?site=bellingcat.com Bellingcat is prime HN, open source intelligence. A lot to learn from. Nothing on Wiki about bias - https://en.wikipedia.org/wiki/Bellingcat https://en.wikipedia.org/wiki/Bellingcat and it often talks about Russian intelligence they'd like to keep quite. Questions why it's flagged on HN are quickly made fuzzy.
- aaron695 6y ago
  OK, I'll call it. I'm going to guess it's these guys with the bot on HN, this bit is only a hunch. On HN I think it's not GPT-3, seems European, and uses some bite size text from the web to create it's output, I assume GPT-3 doesn't get caught doing this. It fits this papers description, but I can't see if they have any example output to be sure past a hunch - https://news.ycombinator.com/item?id=24704952 https://news.ycombinator.com/item?id=24704952

aaron695 6y ago

The one on HN has been running for many months (The one on Reddit only lasted weeks) and has only been called a GPT-3 once I can see. Given it's ability to get a lot of upvotes and no need for a voting ring it would allow a lot of control over discussions through up votes and down votes. Mmmm Russia. It'd be interesting to see it's upvote and downvote history. There are definite attacks on HN you can see play out. I think this bot is actually academic since it is very prolific, it could hide itself a lot more if it wanted to. [edit] OT: I'll give an example of a ongoing Russian attack, Bellingcat is constantly flagged. https://news.ycombinator.com/from?site=bellingcat.com https://news.ycombinator.com/from?site=bellingcat.com Bellingcat is prime HN, open source intelligence. A lot to learn from. Nothing on Wiki about bias - https://en.wikipedia.org/wiki/Bellingcat https://en.wikipedia.org/wiki/Bellingcat and it often talks about Russian intelligence they'd like to keep quite. Questions why it's flagged on HN are quickly made fuzzy.

aaron695 6y ago
OK, I'll call it. I'm going to guess it's these guys with the bot on HN, this bit is only a hunch. On HN I think it's not GPT-3, seems European, and uses some bite size text from the web to create it's output, I assume GPT-3 doesn't get caught doing this. It fits this papers description, but I can't see if they have any example output to be sure past a hunch - https://news.ycombinator.com/item?id=24704952 https://news.ycombinator.com/item?id=24704952