Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > sci.physics.relativity > #667262

Could the AI Persona evolve over time? [With Local AI?] (Was: T-theory A-theory theatheory)

Cross-posted to 3 groups.

Show all headers | View raw

Hi,

With RAG it could react on news. RAG (Retrieval-
Augmented Generation) and other techniques can
make a local AI evolve dynamically — both in knowledge

and personality. Lets say I put such a AI Persona in
a News Group loop. I could feedback the posts there,
to make it also evolve. One can collect community

reactions (upvotes, replies, sentiment). Or apply automatic
evaluation — e.g., sentiment analysis or engagement
score. This becomes your fitness function.

You can even run multiple personas in the same loop
Different viewpoints (e.g., “optimist”, “skeptic”,
“data scientist”) They interact and debate posts.

This is ideal for dead open source projects, such as
SWI-Prolog, to pretend user participation, very important
in our new marketing world that is totally

engagement focus. The quality(*) of content doesn't count.

LoL

Bye

(*) Same for research paper mills. BTW, ChatGPT suggests
me the following tech stack:

LLM: Ollama (LLaMA 3, Mistral, Gemma, etc.)
LangChain / LlamaIndex: context + memory
ChromaDB / SQLite: memory store
Feedparser / API client: ingest news or group posts
Transformers / VADER: sentiment scoring
Simple scheduler or cron job: to run the loop daily

> Hi,
> 
> There are different ways to make an AI Persona.
> For example Twitter aka X, has a grok account.
> But grok is not trimmed towards a particular
> 
> profile, maybe an alter ego of Elon Musk was
> speculated. What could be interesting would be
> to go towards character.ai :
> 
> "Character.ai was established in November 2021.
> The company's co-founders, Noam Shazeer and
> Daniel de Freitas, were both engineers from
> Google. They both worked on AI-related projects:
> Shazeer was a lead author on a paper that
> Business Insider reported in April 2023 "has
> been widely cited as key to today's chatbots",
> and Freitas was the lead designer of an
> experimental AI at Google initially called Meena,
> which later became known as LaMDA.
> https://en.wikipedia.org/wiki/Character.ai
> 
> Maybe this can be even done with Local AI. By for
> example populating a context. Or creating a kind
> of LangChain memory. ChatGPT calls it persistent context.
> 
> So yes you could "spam" an AI Archimedes Plutonium
> probably from your desktop with an AI Laptop, maybe
> already in 2025, or later in 2026.
> 
> bye

Back to sci.physics.relativity | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Re: Taxon (TBox) / Affirm (ABox) was a thing in the 90s? (Was: T-theory A-theory theatheory) Ross Finlayson <ross.a.finlayson@gmail.com> - 2025-11-08 18:21 -0800
  Could the AI Persona evolve over time? [With Local AI?] (Was: T-theory A-theory theatheory) Mild Shock <janburse@fastmail.fm> - 2025-11-13 18:00 +0100
    Boris the Loris and Nazi-Retard Julio (Was: Could the AI Persona evolve over time? [With Local AI?]) Mild Shock <janburse@fastmail.fm> - 2025-11-13 18:07 +0100

csiph-web