Inspiration

Our inspiration came from recent online discussions of the carbon footprint AI leaves behind. Many people use it for everything now and may not be aware of how much energy it actually costs to perform a request. We hope to make users mindful of using Artificial Intelligence to reduce environmental impact.

What it does

As you start typing in the ChatGPT prompt box, EcoGPT will detect if your query is complex or simple. If it is simple, it will display a box with a reminder that performing a search on google will reduce the environmental impact and provides resources for other ways to complete the search. Through the EcoGPT Dashboard you can view how many requests you have made and the impact of those requests.

How we built it

We built our project by creating a model that classifies the user prompt as simple or complex. A complex query would allow the user to go ahead and use ChatGPT. A simple one would show alternative ways to get the result such as a Google Search. This is provided as a hyperlink for ease of use and accessibility.

We designed an API hosted in the cloud that the chrome extension connects to. The API analyzes the user query using our model and calculates the carbon footprint of ChatGPT requests. It is designed in Flask using a GraphQL design to implement client-server communication and deployed to render.

The chrome extension injects a modal to the ChatGPT interface with our EcoGPT Assistant along with an easily accessible analytics dashboard.

Challenges we ran into

Most of our team has not had experience with building Chrome Extensions and since ChatGPT dynamically allocated elements, we had to use alternate ways to grab elements from the DOM.

Accomplishments that we're proud of

We got a good understanding of how chrome extensions work and are developed, with each script having a unique purpose. We also worked well doing pair programming and working through challenging problems without losing motivation.

What we learned

We learned about the under the hood of how chrome extensions work and Interact with webpages and api's. We learned how to develop an API using GraphQL, and host it within the cloud. Additionally, we were able to recognize the importance of a positive mindset and encouragement for each other to persist through any obstacles; which ultimately lead to a finished product. This taught us that even if an issue seems impossible to solve, it is not.

What's next for EcoGPT

We want to train a machine learning model to detect if each query is simple or complex.

Tech Domain

thinkbeforeyouprompt.tech

Share this project:

Updates