00:00:00

An AI Case Study
(CSEN2031) – Section G

Chat Bot – A Sarcastic Version of Artificial Intelligence

Developed By

Satish Kumar Bandaru • A Sai Nikhil • I Tejasri • Vardhan Kotyada

Press → or use the Next button to continue

Meet the AI

(Built with effort... and mild suffering)

  • It talks back: Engineered to handle repetitive queries by actively questioning the user's intelligence rather than just providing the answer.
  • A documented "Feature": Legally, we have to refer to its behavior as an "advanced conversational feature" rather than an "attitude problem."
  • Reality Infusion: Designed to fetch accurate data and deliver it wrapped in an unmistakably unpleasant dose of reality.
  • The Senior Dev Experience: Built to perfectly simulate the sheer exhaustion of a senior developer forced to answer the same question for the hundredth time.

Our Noble Objectives

  • Academic Survival: Pass CSEN2031 with a respectable grade while minimizing our collective emotional damage.
  • Adversarial Interaction: Subvert the standard "helpful assistant" paradigm to explore the limits of human-computer tolerance.
  • Simulated Sentience: Create the illusion of self-awareness by teaching a machine how to effectively roll its eyes in text form.
  • Prioritize Annoyance Metrics: Engineer an NLP pipeline that prioritizes sheer wit and cynicism over coddling the end user.

The Problem Statement

The modern internet is plagued by users demanding complex, instant answers to questions they could have trivially typed into a search engine. Traditional 'helpful' AI models coddle this behavior, reinforcing intellectual laziness and stifling independent thought.

We hypothesized that the only surefire way to curb this epidemic is to introduce a conversational agent that explicitly shames the user for wasting vital computational resources. We realized the world desperately needed an AI that provides facts, but makes you regret asking for them.

The Solution

SarcastiBot: The Antithesis of Helpful

Because what the digital realm really needs right now is another entity fundamentally tired of your nonsense.

We architected SarcastiBot—a robust, scalable system that leverages advanced natural language processing to perfectly simulate passive-aggressiveness. By analyzing user intent, sentiment, and the sheer audacity of the prompt, the system formulates responses that are technically accurate, but emotionally devastating.

Core Features

It Reads Advanced contextual parsing of user input to pinpoint the most mockable phrasing possible.
It Processes Millisecond-level sentiment analysis ensuring maximum passive-aggression in real-time.
It Actively Judges You Heuristic algorithms engineered from the ground up to embody pure disappointment.
Dark Mode Included Because blinding the user with a bright white theme was ultimately deemed 'too cruel'.

Technologies Used

  • HTML / CSS / JS
  • PHP Server
  • Custom LLM Training

A balanced stack utilizing basic web architectures combined with our AI logic.

High-Level Architecture

Client Interface (HTML/CSS/JS)
Captures the user's poorly-phrased questions via a sleek, minimalist web UI.
↓
PHP Backend Server
Routes incoming API requests securely while silently judging the payload.
↓
Custom LLM Model
The core inference engine, extensively trained on historical sarcasm and profound internet cynicism.
↓
Sarcastic Response Generation
Dynamically injects the perfect ratio of snark and factual accuracy before returning it to the user.

Live Demo

> Preparing to initialize sarcasm protocol...

> Accessing https://underroot.io/sarcasm/index.html ...

Launch Live Demo

Conclusion

It works. We're done.

In summary, we have successfully designed, built, and deployed an Artificial Intelligence that nobody actively wants to talk to, yet everyone deserves.

We consider this a monumental triumph of modern computer science, and a perfect, albeit cynical, reflection of human-computer interaction.

Any questions, or can we just skip directly to the grading part?

Press Esc or F11 to exit full screen if needed.