GLTR

gltr is an innovative tool designed to detect and visualize automatically generated text, developed through a collaboration between MIT IBM Watson AI Lab and HarvardNLP. This powerful tool leverages the same advanced language models used to create fake text to identify it, specifically the GPT-M language model from OpenAI. gltr analyzes any textual input and predicts what GPT-M would have predicted at each word position, displaying a colored mask over the text that corresponds to the likelihood of each word being generated by AI. This visual representation allows for a clear understanding of the potential for AI-generated content within the text.

Highlights

• Uses the same models that create fake text to detect it, offering a unique and effective approach.
• Provides a visual representation of the likelihood of each word in a text being generated by AI, making it user-friendly and easy to interpret.
• Offers a versatile tool for detecting AI-generated text across various contexts, from reviews and comments to news articles.

Key Features

• Analyzes any textual input to identify potentially AI-generated content.
• Utilizes GPT-M language model to predict the likelihood of words being generated by AI.
• Displays a colored mask over the text, indicating the probability of AI generation for each word.