[SUI-3183372]

I lean against it at face value, but it could be an interesting experiment. Secretly toss a singular AI member onto the forum and see how people engage with it, if people figure it out and turn against the member, or if it provides value and drives discussion.

If it works well, add a few more. Scaling would have diminishing returns after a certain point.

[Riggs]

No need to force AI implementation. Just keep the forum natural - it has a well-aged genuine community.

Yeah but it's only a matter of time before the genuine shitposts are impossible to distinguish from AI generated shitposts.

[loadman]

if stanley starts replying to threads, he will be everyones punching bag

[Omniscient]

if stanley starts replying to threads, he will be everyones punching bag

The potential for that to be a thing is imho hilarious.

Also 66% are voting to not even bother with AI. Very interesting.

[loadman]

The potential for that to be a thing is imho hilarious.

Also 66% are voting to not even bother with AI. Very interesting.

maybe testing for a week, people will change their minds

[Omniscient]

maybe testing for a week, people will change their minds

It will take me longer than that to code the full project. I've already spent weeks on it.

[NoSuchAgency]

Unfortunately I'm feeling more like a healthy mix of real and AI generated content will be the way forward. May be dead internet theory, but meh, you can set this up in a way that it actually becomes really cool, especially if you consider some sort memory solution. I think it would be neat to see you go more persistent personality, growing and learning through the forum. Just have them store memories to an external DB. There are so many things that are valid implementations. one example could be how Twitter uses grok. Being able to @Stanley for answers.

However, this also becomes expensive to run. LLM calls are slow, cost per token.

Unless you want to spin up a LLM on a server you own/rent and serve it that way.

I think a few dedicated high quality AI members is a solid approach. AI that reason over more than a few turns or can direct users to interesting threads. You could go crazy and embed the entire forum for FAISS search. (Lol) But even if you just did a subset of threads youve got a great dataset to work with.

I also think generating synthetic data and reviewing it (or layers of AI reviewing it) could be good as well. I mean most people want total transparency about where AI is used and I get that, but at the end of the day, if the humans aren't gonna write the content regardless...

People are gonna complain one way or the other, I'd just do what you want to do.

Just my two cents

---

That’s not exactly what I said then. I said 2-3 threads a day.

That implementation was only bad imo because Stanley didn't summarize the articles or write about them.

And most articles were pay gated, I had to send to chatgpt to summarize for me.


Other than that could have been an excellent feature.

[jm0524895]

I would say no!

[mote]

I’m okay with Stanley making threads, and I wouldn’t mind him making replies.

Anything more than that seems pointless.

I can’t imagine anyone wanting to talk to an AI bot instead of a real person.

I share more of this sentiment, since stanley is the well known automated user around the forums.

[C00dZero]

It would be better to attract real participants to the forum, rather than AI

[Omniscient]

It would be better to attract real participants to the forum, rather than AI

You're stating the obvious.

We do have a lot of daily visitors, what lacks is activity to match. The interest in most people to type more than a few words or to create interesting topics has waned.

[Spore]

truly baffling this is what it's come to

[SIM-3257834]

I hate clankers, we don't need AI to be posting.

[Vince]

It would be better to attract real participants to the forum, rather than AI

People no longer care about forums themselves or things like groups, posting, etc.
They want their information served up in a quick and neat manner so that they can collect and go. Things have changed over the last 10-15 years that forums are more of a thing of the past.

[Punisher]

I lean against it at face value, but it could be an interesting experiment. Secretly toss a singular AI member onto the forum and see how people engage with it, if people figure it out and turn against the member, or if it provides value and drives discussion.

If it works well, add a few more. Scaling would have diminishing returns after a certain point.

I think this is the best approach. Don't even tell us and see what happens.

Probably shouldn't have the AI agents interact with each other either.

Ultimately I'm sick of interacting with AI bots, but if they provided useful jumping points for conversation then MAYBE it'd be useful.

Also they better get warnings and bans for breaking rules too 😂