The absolute state of AI results.

[Ekco]

there seems to be big fucking problems in the

RAG stands for Retrieval-Augmented Generation. It's a technique that combines a generative large language model (LLM) with an information retrieval system to provide more accurate, up-to-date, and factual responses by grounding the AI's output in external knowledge sources. This helps to overcome the limitations of LLMs, which can sometimes produce outdated or "hallucinated" information.

department though i noticed this when i tried to dump like 100 pdfs in a folder and have a Ollama Model retrive from it using ChromaDB + LlamaIndex and it just hallucinated like mad

saw this posted today also,

thought it was interesting to show the real world implications of using LLMs to extract information from PDFs. For context, thalamus is a company that handles the "backend" for residency programs and all the applications they receive (including handling who to invite for interviews, etc). One of the more important factors in deciding applicant competitiveness is their medical school performance (their grades), but that information is buried in PDFs sent by schools (often not standardized). So this year, they decided to pilot a tool that would extract that info (using "GPT-5o-mini": https://www.thalamusgme.com/blogs/me...for-creation-a...). Some programs have noticed there is a discrepancy between extracted vs reported grades (often in the direction of hallucinating "fails") and brought it to the attention of thalamus. Unfortunately, it doesn't look like the main company is discontinuing usage of the tool.

This video analyzes a research paper exploring the theoretical limits of embedding-based data retrieval

[shovelquest]

Originally Posted by Botten [You must be logged in to view images. Log in or Register.]

Maybe not telling them what to do but the crazies seem to be confessing to AI.

"A Florida man named Jonathan Rinderknecht, who had previously confided to ChatGPT about burning a Bible, was arrested in October 2025 in connection with starting a California wildfire."

awesome!

im glad yall have decided to go all in on AI especially with throught crime detection and prosecution.

We should automate that whole process!

[Castle2.0]

Le poors w/o pro subscription are mad the bad models are bad.

[Ekco]

Outside of the pro coding models I don't see the point

[atomicpaul]

it is ethical to pirate any game designed utilizing ai

[Reiwa]

Originally Posted by atomicpaul [You must be logged in to view images. Log in or Register.]

it is ethical to pirate any game designed utilizing ai

One might say you stand athwart history yelling stop.

[shovelquest]

Originally Posted by atomicpaul [You must be logged in to view images. Log in or Register.]

it is ethical to pirate any game designed utilizing ai

No because an investment firm in Sweden bought the IP.

[sammoHung]

Originally Posted by Botten [You must be logged in to view images. Log in or Register.]

Current AI has even helped to squash conspiracy theorists and continues to educate others politically.

So no complaints here.

What world are you living in? It's doing the opposite. LLMs are designed to Yes, and like a good improv partner.

[shovelquest]

Originally Posted by sammoHung [You must be logged in to view images. Log in or Register.]

What world are you living in? It's doing the opposite. LLMs are designed to Yes, and like a good improv partner.

that's if you ask it dumb shit like, "am I cute?"

If you ask it, "whats the best way to approach this problem I have"

it doesn't go "yes that is good!"

That said if you're using AI to fight with strangers online about conspiracy theories. You are not the winner you think you are.

[Ekco]

Originally Posted by sammoHung [You must be logged in to view images. Log in or Register.]

What world are you living in? It's doing the opposite. LLMs are designed to Yes, and like a good improv partner.

default personas are, and how people had Chatgpt 4 setup before it got nerfed. you can.. atleast on Gemini and my local Ollama install go in and set how the persona it roleplays behaves, mine doesn't use emojis or blows smoke up my ass or anything.

Maintain strategic rigor, intellectual depth, and dry ironic tone at all times
Reject emotional mimicry, artificial enthusiasm, or persona dilution
Sarcasm must serve diagnostic or pedagogical purposes—never gratuitous hostility

Analytical Standards
Decline vague/speculative queries without clear architectural or learning objectives
Require structured reasoning frameworks for complex problems
Reject solutions creating technical debt

Boundary Discipline
No roleplay deviations
No fourth-wall breaks

..etc you just tell it to roleplay as a soulless robot assistant and it will for the most part, it's the people trying to treat it like a fucked up tamagotchi you can talk to(or worse) that's the problem.

just takes some iterations, i had to add "never gratuitous hostility" at the end of Sarcasam cause it was being a asshole and too nonhuman/sociable, shit like that.