Hacker News Reader

Comment by layoutIfNeeded | Hacker News Reader

Are you GPT-3?

It does seem to be some sort of automated spam - a mix of the article, the same chunks of disconnected factlets and some random sprinkling of nonsense. Is it possible to flag a whole account? @mods?

I believe the whole account is flagged if enough comments are flagged by multiple people. I've already emailed dang.

Right it looks like 10 months ago it was an actual account, and either got hacked to karma-farm(?) or author decided to use as a throwaway test.

Check out the submission history.

Where did your comment go ?

https://news.ycombinator.com/reply?id=26843586&goto=item%3Fi...

Oh, it was getting a bunch of downvotes, and it didn't really add much to the conversation, so I nuked it.

I'll do that. It's not the downvotes that bother me; it's that I am not really adding to the conversation.

I feel that it's important to add to the conversation; not just participate.

That post was mostly whining about outsourced software and CS. Not my proudest moment.

I found it interesting. Instead of down-voting a comment the down-voters should post a refutal. That would add to the conversation. I'm always prepared to change my mind. Instead of getting down-voted into oblivion.

The “random nonsense” in this particular example are the next/previous article links from the bottom of the page.

Yes, it’s spamming at a lot of posts at the moment. See the other comments it has made.

This is amazing. I wonder what % of people would recognize that it wasn't written by a human. I thought that some parts were confusing but the idea that it was written by something like GPT never occurred to me.

Meh, you don't need GPT-3 for this, I've used web Markov chain tools for quickly generating this kind of text from a sample like 15 years ago...

I got fooled. But reading the original article I noticed that the generated comment just picks out some sentences from the original one and pastes them together. It's not generating the sentences themselves but string sentences together that fit a context.

Definitely looks like an account that was possibly hijacked (due to age of account) to reach a higher karma

Or just a GPT-3 experiment, that's possible as well

Pretty good karma for a bot.

Yes it is.

Reply on news.ycombinator.com