This project was build as part of the Cohere AI 48-hour Hackathon #3

Introduction

• Debait is a Conversational AI designed to debate about an article/topic of your choice.
• It uses Co:here’s Generate and Classify models
• With a bit of prompt engineering, history modelling and classification we can obtain a very powerful output from Co:here’s model!

Getting Started

1. Clone this repo
2. Add your API KEY into /config/config.env from co:here. Read the docs here for help.
3. Create an environment on Python 3 and install the packages from the requirements.txt file
4. To start the application run main.py

Co:here and LabLab

• Find out more about co:here.
• My LabLab profile.

Building Debait (model framework and GUI mock-up)

Applications of Debait

1. Fake News – Help prevent polarising users through social media exposure
2. Education – Help develop ideas, expand students’ critical thinking skills, articulation, research skills, and resolution of conflict.
3. Content Generation for Media - Articles, magazines and blogs can use Debait to generate content and ideas.

Improvements

1. Better GUI and Faster AI Response
2. Fine-tuning the Co:here Generate model and using more data for Co:here Classify
3. Allow user to control the tone and length of the response from the AI
4. Show more metrics with response to move towards explainable AI

Debait in action