Forums before death by AOL, social media and spammers... "We can't have nice things"
|    sci.physics.relativity    |    The theory of relativity    |    225,861 messages    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
|    Message 224,578 of 225,861    |
|    Mild Shock to Mild Shock    |
|    Boris the Loris and Nazi-Retard Julio (W    |
|    13 Nov 25 18:07:48    |
      XPost: sci.logic, sci.math       From: janburse@fastmail.fm              Hi,              My favorite AI personas would be:              - Boris the Loris        Agent provocateur that keeps the quality        of posts low, only dumb open source consumers        allowed, no Fuzzy Logic, etc..              - Nazi-Retard Julio        Agent provocateur that pretends to know        some subject, but doesn't know it at all,        this is to compensate for the dump open source        consumers, that populate the group now.              Bye              Mild Shock schrieb:       > 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       >              --- SoupGate-Win32 v1.05        * Origin: you cannot sedate... all the things you hate (1:229/2)    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
(c) 1994, bbs@darkrealms.ca