/RAI-AI_Detector

An AI message detector written in python using tensorflow.

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

Several parts of this README file have been written by AI

you can test the AI by figuring out which parts were.


AI Detection Program

This is an AI detection program written in Python using TensorFlow, designed to analyze and classify messages from humans and ChatGPT. The program utilizes a dataset filled with Discord messages and ChatGPT-generated messages to train a machine learning model for identifying AI-generated content.

dont put your mouse here.

Capabilities

Capabilities Y/N
Detect if text is written by a human or not ✅
Differentiate between news articles an AI ✅
Differentiate formal text for both human and AI ✅
Differentiate between other chat bots ❌
Differentiate between government text and AI ❌
Accurately classify short strings of text ❌
Point out strings of text that aren't completely AI ❌
Detect strings of text in languages that aren't English ❌

Requirements

To run this program, you need the following:

  • Python 3.x

  • TensorFlow library

  • Numpy library

  • Pandas library

  • Scikit-learn library

You can install the required libraries using pip:

pip install tensorflow numpy pandas scikit-learn

Dataset

The dataset used for training this AI detection program consists of Discord messages from the "E - D G Y" Discord server and ChatGPT-generated messages. It is crucial to have a diverse and representative dataset to ensure the effectiveness of the detection model.

The dataset should be organized in a CSV (Comma-Separated Values) format, with each row representing a message and containing the following columns:

  • class: Classification of the string (0 = Human, 1 = AI).

  • message: The labeled string of text.

AI traning dataset example

Model Training

  1. Preprocessing: The dataset should be preprocessed to remove unnecessary characters, clean the text, and convert it into a suitable format for model training. The messages cannot contain commas or quotations due to the CSV file format.

  2. Model Training: After modifying the dataset CSV files to your needs, you may then run the train.py file and let it run the epochs. After the training is over it will ask you for user input and will then classify the message as AI or human generated.

yes im hosting these on discord, kill me.

Usage

To use the AI detection program, follow these steps:

  1. Ensure that you have installed all the required libraries mentioned in the "Requirements" section.

  2. Obtain or create a suitable dataset containing Discord messages and ChatGPT-generated messages in CSV format. Make sure it includes the necessary columns: class and message.

  3. Run the train.py script to train the AI detection model. This will save the trained model to a file.

  4. Once the model is trained, you can use it for detecting AI-generated content. Modify the main.py script to load the trained model and provide new messages as input. The script will predict whether the input messages is written by a human or ChatGPT.

python main.py

Future Improvements

  • Increase dataset diversity: Gather a more diverse dataset with a wider range of human messages and ChatGPT-generated text to improve the model's accuracy and generalization.

  • Train it on other languages including programming languages to improve classification.

  • Make a headless mode for the AI to allow automation.

  • Prevent it from detecting the American constitution as AI.