Ask HN: Posting using ChatGPT should be banned?

2 points by nerder92 3y ago ↗ HN

I've recently seen the news about stackoverflow banning ChatGPT generated answers, and also HN itself seems to ban this kind of questions/posts. What do you think of that?

To me is kinda similar to fair use in copyrighted material, for instance using some parts of generated content and some original parts in one answer/post shouldn't be banned. Also with which degree of confidence we are able to detect if the content is actually generated by AI or not?

It's quite an interesting topic on how this will change moderation and if this is actually possible to do on scale

Couple of opinons from HN moderator will be highly appreciated here!

10 comments

[ 3.2 ms ] story [ 29.0 ms ] thread

smt88 3y ago ↗

It's a moot point because you can't ban ChatGPT anywhere. There's no way to reliably tell the difference between AI output and a human who doesn't speak English as a first language (or is strange or has a cognitive disability).

derivagral 3y ago ↗

Personally, chatGPT-only posts fall under "low effort". If the post involves something more than a few paragraphs tailed by 'written by chatgpt' I would find that different, but so far most seem to be the former.

Trouble_007 3y ago ↗

>Should HN ban ChatGPT/generated responses? - [ 7 hours ago - 51 comments ] : https://news.ycombinator.com/item?id=33945628

>Tell HN: Copying and pasting from ChatGPT unsolicited sucks - [ 4 days ago - 188 comments ] : https://news.ycombinator.com/item?id=33891538

>Stack Overflow bans ChatGPT as 'substantially harmful' - [ 5 days ago - 31 comments ] : https://news.ycombinator.com/item?id=33882298

>Use of ChatGPT generated text for posts on Stack Overflow is temporarily banned - [ 6 days ago - 140 comments ] : https://news.ycombinator.com/item?id=33861846

Trouble_007 3y ago ↗

https://stylometry.net/user?username=clay-dreidels

>User not found. Either this user does not exist,

  they have not written at least 10,000 characters, or their account was created after the day I scraped the posts.*

I was hoping that the database was updated by one of the unofficial RSS feed. As per: https://hnrss.github.io/

Edit to add;

https://news.ycombinator.com/threads?id=clay-dreidels

Thanks Mods

hxugufjfjf 3y ago ↗

Show HN: UserScript to detect GPT generated comments on Hacker News https://news.ycombinator.com/item?id=33906712

inphovore 3y ago ↗

It’s kind of nice for laughs. Maybe there needs to be a new button “[] suspiciously AI” we could click it under suspicion and see how the stats resolve.

ggm 3y ago ↗

I agree, but I have also been (falsely) accused of doing it. the current SOTA in detectors is arguably not good and promotes a spurious confidence interval.

It's my belief if you are in the training set, or write with quirks, you have a high chance of being a false positive.

WheelsAtLarge 3y ago ↗

It's a "Genie out of the bottle" thing. It can't be stopped. It will eventually be part of people's workflow in the same manner search engines are part of most people's daily life. Banned or not banned people will use it. Arguing about it won't change a thing.

My concern is that it will become a rich person's toy since eventually OpenAI will charge for it.

cloudking 3y ago ↗

I made a post asking if AI content should come with a disclosure: https://news.ycombinator.com/item?id=33858508

Personally I think it's okay if it's disclosed, users can decide to weight it accordingly with their votes.

johlits 3y ago ↗

Thing is, we are all bots trained on our experiences throughout life. How are we different from the bots? I believe we must accept and learn to live with the bots from now on.