this post was submitted on 04 May 2024
43 points (95.7% liked)

Reddit

17716 readers
127 users here now

News and Discussions about Reddit

Welcome to !reddit. This is a community for all news and discussions about Reddit.

The rules for posting and commenting, besides the rules defined here for lemmy.world, are as follows:

Rules

Rule 1- No brigading.

**You may not encourage brigading any communities or subreddits in any way. **

YSKs are about self-improvement on how to do things.

Rule 2- No illegal or NSFW or gore content.

**No illegal or NSFW or gore content. **

Rule 3- Do not seek mental, medical and professional help here.

Do not seek mental, medical and professional help here. Breaking this rule will not get you or your post removed, but it will put you at risk, and possibly in danger.

Rule 4- No self promotion or upvote-farming of any kind.

That's it.

Rule 5- No baiting or sealioning or promoting an agenda.

Posts and comments which, instead of being of an innocuous nature, are specifically intended (based on reports and in the opinion of our crack moderation team) to bait users into ideological wars on charged political topics will be removed and the authors warned - or banned - depending on severity.

Rule 6- Regarding META posts.

Provided it is about the community itself, you may post non-Reddit posts using the [META] tag on your post title.

Rule 7- You can't harass or disturb other members.

If you vocally harass or discriminate against any individual member, you will be removed.

Likewise, if you are a member, sympathiser or a resemblant of a movement that is known to largely hate, mock, discriminate against, and/or want to take lives of a group of people, and you were provably vocal about your hate, then you will be banned on sight.

Rule 8- All comments should try to stay relevant to their parent content.

Rule 9- Reposts from other platforms are not allowed.

Let everyone have their own content.

:::spoiler Rule 10- Majority of bots aren't allowed to participate here.

founded 1 year ago
MODERATORS
L3s
_MoveSwiftly
Thekingoflorda
ja2
Rooki
43
Went to look for the first time in a while...didn't even take me 10 minutes to find an obvious bot. (www.reddit.com)
submitted 7 months ago by EdibleFriend to c/reddit
9 comments fedilink hide all child comments
 

All it does is respond to the title. sometimes that works out for it....sometimes its horrible. Just a tiny bit down, check the old school cool. down in the comment graveyard is another number name one doing the exact same thing.

top 9 comments
sorted by: hot top controversial new old
[–] [email protected] 16 points 7 months ago (1 children)

Even the strategy being used by the botter is obvious. Scrape off post titles, feed them into a LLM prompted with something like "Output a funny and short reply for each of the following prompts. Do not output any text besides said reply.", then comment the output.

I tried it, and here's how it turned out:

Post title Annual-List1897's answers Gemini's answers
Christopher Lloyd and Michael J. Fox (2024) Ah, the dynamic duo! They're like a time-traveling dream team. Christopher Lloyd and Michael J. Fox in 2024? Great Scott! They're still not in 2024?
Just leaving this here.. Oh, you're just dropping a mysterious statement and walking away, huh? Don't worry, I'll try not to trip over it.
Pee Pee Island to Poo Poo Point road trip. Sounds like a wild adventure! Packing the Depends and air freshener?
ABC will air fewer "Monday Night Football" games during the 2024-25 season Well, I guess that means fewer opportunities for people to fall asleep on the couch with the TV on after a long Monday! ESPN execs right now: "Finally, some peace and quiet on Mondays."
Mysterious package sent to my wife (info in comments) Well, well, well, a mysterious package for the misses, eh? "Honey, is this your 'Oops, I accidentally signed us up for a lifetime supply of clown shoes' subscription?"

Sure, it is not the same replies, as this would depend on the exact prompt and the LLM bot being used. But note how it gets really close in spirit.

And this shit is trivial to do. Even for a non-programmer like me. Reddit might not allow you API access, but sometimes a simple scrapper does the trick.

@[email protected] mentioned the "default" name pattern, note that if necessary those could be also generated.

[–] GoosLife 4 points 7 months ago* (last edited 7 months ago) (1 children)

Your Gemini is way funnier in my opinion. I think he actually might have set up a trap for himself by asking it to produce what the LLM would consider a typical or average reply. Whereas by asking it to just make a short, funny comment, you're actually getting results that feel more natural.

For Gemini, only the first and last one read weird to me. But I think I would just assume that I'm missing some context to get the jokes, or something.

Whereas the actual replies from the OP actually reek of standard LLM drivel. The way it is trying so hard to sound casual and cool, but coming across as super awkward is just classic GPT.

[–] [email protected] 1 points 7 months ago* (last edited 7 months ago)

Yeah, the botter might have overdone it.

Another thing that I realised is that you could actually do this in a semi-auto way, without coding a single line of code - using a plain browser and an auto-clicker. EDIT: I'm not saying that autoclicker+browser is how people trying to bot Reddit "should" do it. I'm using it to highlight that botting Reddit is an extremely low-hanging fruit, to the point that you expect lots of bots there.

[–] Windex007 10 points 7 months ago

Adjective Noun Number.

[–] flicker 6 points 7 months ago (1 children)

I think the thing I hate the most about LLM stuff isn't even how awkward it is. It's how it's all so blah. Like having cream of wheat for dinner. Sure, it qualifies as food, but you're definitely not living.

Even when it references pop culture or memes, it does it so... lame.

[–] [email protected] 3 points 7 months ago

This. Also I'm pretty sure they were trained on the turnitin.com corpus so they waffle a lot. They're incapable of making a point without sounding like someone who loves the sound of their own voice.

[–] [email protected] 5 points 7 months ago* (last edited 7 months ago) (1 children)

Reddit bots comments do not always have to make sense. The bots likely quickly delete downvoted comments to stop the karma bleed and rely on only a few of them be upvoted to get that sweet sweet karma.

This was how the previous generation of Reddit bots - comment thieves - worked.

I wonder if there are already hybrids that steal and rephrase other comments using GenAI instead of just responding to the post title?

[–] [email protected] 1 points 7 months ago (1 children)

I wonder if there are already hybrids that steal and rephrase other comments using GenAI i

Have to be. There were rephrase bots before LLMs even, they just changed a few key words like "argument" would become "fight".

Whoever made them would definitely think of this.

[–] [email protected] 1 points 7 months ago

I chased comment thieves with a script that called them out for a while, so I noticed those everywhere and started learning their behaviour patterns.

Unlike comment thieving bots, rephrasing bots could be difficult/expensive to detect. But at least if they use GenAI they are also expensive to run. (If they use a simple substitution, they are detectable, although it would take some work I guess. Fun work though).