Exploring a New Idea: Automating Content Generation for 12 Months with AI-Enhanced Characters

[wpraiz (Jose Icaro)]

Hey everyone,

I’ve been thinking about introducing a fresh approach to content creation on the forum. The idea is to leverage the power of AI to generate uplifting content for a period of 12 months. Imagine having 50-100 AI-driven characters that interact daily, engaging in discussions, and creating content that’s focused on positivity and happiness. This could be a wonderful departure from the usual news of corruption and tragedies that tend to dominate.

The generated content wouldn’t stop there. Once we’ve accumulated a wealth of positive and valuable insights, we could consider donating these creations to various causes and organizations, making a difference in the lives of others.

Currently, we have 110 registered human users, primarily consisting of students from my startup and other existing projects. Over the past four months, if we’ve found that there haven’t been many new posts or contributions, I believe it might be more beneficial for me to invest my time in honing my Python skills, creating stories for my son, and producing content for those who are in need. The content I plan to create will include audio, images, and text in the form of my multimodal articles.

This post marks my first contribution to the community, and I hold forums in high regard. Back in the country where I was born and currently reside, it was through forums that I absorbed valuable content and made friends who have stayed with me to this day. Hence, I’d like to kindly request permission and guidance. The Discourse platform caught my eye immediately, and it’s truly a pleasure to engage with content that challenges me enough to post.

Having left my computer science studies a decade ago, despite my experience in programming with languages like C and more, I’ve spent around three hours today attempting to update the Discourse instance I’ve set up. Upon completing the task, I was taken aback by the lack of engagement from the community members. In Brazil, there’s yet to be a similar concept, despite the existence of beautiful initiatives from nearby communities. I believe it’s important to gather sincere opinions from the official Discourse community, and that’s why I’m reaching out.

Your insights and thoughts are immensely valued as we navigate this endeavor. I’m looking forward to hearing from you all.

Warm regards,

[Ed_S]

This sounds like something you could dry-run in a matter of hours or days - the AI contributions could be accelerated, and you would learn how the contributions interact with each other.

But you need to know where your initial inputs are coming from: what causes an AI to start a new thread? And what topic will it address?

Your setup, in any case, is very different from mine. My forums are focussed on a topic - a hobby, interest, or activity. There’s no place for commentary on current affairs, politics, news, tragedies, or anything of general interest. In my experience, when a tightly-focused community allows off-topic contributions, it goes off the rails. A smallish number of vocal contributors will use it as a soapbox to share their personal opinions and experiences, and arguments will follow.

If I enjoy skateboarding, food, and travel (for example) then I need three separate forums to indulge those three different interests. It does no good to mix them up, because two-thirds of my topics will be boring (or controversial) to any particular reader.

It may be that among young people - you mention students - everything is interesting and everyone shares the same interests, and there is no such difficulty. But it might be more likely that the most “engaging” interests are crowding out the minority interests. The ones which win, in volume, are the ones which can support argumentation - politics, gossip, news. Negativity wins, because argument causes people to want to respond. The problem being that it doesn’t make them happy.

[maiki]

@wpraiz

Imagine having 50-100 AI-driven characters that interact daily, engaging in discussions, and creating content that’s focused on positivity and happiness.

I’ve been experimenting with this: I’ve been interacting with ChatGPT as various Animal Crossing characters. My family plays Animal Crossing, so I wanted to see if I could make chatbot villagers.

I’ve been able to do it fairly well one-on-one, where I create appropriate prompts to interact with, but ultimately I wanted to make a little hobby kit for people to play around with.

While there are a lot of moving parts, I wanted to share what I’m doing at a high level, in case folks are doing similar.

Basically, each villager/character is a Persona, and each message they send is short like in the game…

…and I track all the messages they send in the forums, as what I call “indoles”:

Etymology 2

From Latin indolēs (“inborn quality, nature”), from indu- (“within, in”) + ol- (“to grow”) (an affix also found in abolish and adolescent).

Pronunciation

• (Received Pronunciation) IPA(key): /ˈɪndəʊ̆liːz/

Noun

indoles (uncountable)

1. Natural disposition; innate character; unalterable intrinsic traits and qualities (collectively).[1]

References

1. ^ The Oxford English Dictionary (2007)

Okay, so here is what I do:

I create dozens of indoles for each character, made from quotes from the game. And I send a bunch of them as part of the system prompt to the chat API. From the Common Crawl, ChatGPT is already effective at pretending to be most characters, but this way I’m able to interject “personality” in each API call.

Okay, here’s the fun part! Each response that goes back into my forums becomes a new indoles!

When I filter the indoles to go along with the chat API call, half of them are recent responses, and the rest are randomly chosen from the rest (I put a limit on how many are included, so I don’t mess up my API calls).

This allows the Persona to slowly over time become “personalized” based on the conversations they’ve had. And because I load recent responses, there is a bit of “awareness” of the Personas interactions.

I haven’t quite gotten to it yet, but eventually I plan to have them talk to each other, but I’m rate-limiting responses (for lots of reasons: ease, funding, because I like relaxed forums), and I haven’t quite figured out the @-mention logic for them yet.

---

Let us know if you are building something similar.

[wpraiz (Jose Icaro)]

Thank you, Ed.

I find it extremely important to outline the operational aspects and decision-making, but initially, I thought of something more simplified with around 10 bots, where 2 have ‘interlocution’ characteristics, and the other users with basic frequencies of 1-2 days with few posts and replies.

I will probably use some behavioral and personality studies (haven’t decided which ones yet, but I’ve already gathered several) as a basis for creating the ‘characters,’ analyzing patterns and response times with simple scenarios like ‘response length,’ ‘potential for questions,’ ‘humor score,’ 'personality or behavior/habit of researching before asking or answering.

[wpraiz (Jose Icaro)]

Absolutely fantastic!

I embraced the initial idea because since September of last year, I’ve been creating and adapting codes and automations so that my 4-year-old son can interact, request images, engage in voice interactions, and have fun with me.

For this project, I also have an idea related to translating indigenous languages using GPT (yanomai - an open-source project). As part of this endeavor, I plan to include characters or contexts inspired by figures from Latin American indigenous cultures and folklore. This serves not only to diversify the content but also to create e-books that can be donated in the future.

I would greatly appreciate suggestions for creating stops/checkpoints with prompts like policy or argumentative modulations to guide conversations. Initially, I’m planning to start with around 10 characters for Ed, without the need for a database or machine learning.

Simple mermaid