Original Work
Energy Saver AI - Prompt Optimizer
An AI-driven application that analyzes and optimizes user prompts to reduce computational load and energy consumption.
Built with T5 transformer model, measures energy savings using Zeus library, and includes web interface for accessibility.
Tools & Technologies
Original Work Proposal
Project Goal: For my original work, I plan to produce a website or app with Chrome extension capabilities that can help reduce AI energy consumption. This AI model will optimize prompts that are plugged into models like GPT, and by optimizing these prompts, the energy used by GPT will be greatly reduced.
My AI model would take in a person's prompt, reduce the amount of text, and combine with similar prompts and output a more energy efficient prompt with the stats of how much energy would be saved. An app like this would not only reduce energy consumption directly by reducing prompt size in multiple aspects, but it would also raise awareness of just how much energy one word causes these models to exert.
Also, by providing the same response for multiple similar prompts plugged in, the AI model would need to exert much less energy on creating a completely new response. I will utilize machine learning and deep learning alongside other algorithms to create an AI model with this prompt optimization abilities.
Project Impact
This work would help the environment by reducing energy wasted and society in general by teaching them how to prompt more efficiently to reduce the energy they waste. People's carbon footprint from using AI would be reduced drastically just by using a different AI model to write more efficient prompts.
By working in this unexplored route of prompt optimization, it could also lead to large tech companies researching similar solutions to make their AI more efficient. I want to further a career in AI specifically researching about Green AI, and this project would allow me to learn a lot more about just that.
Display Board
This display board was presented at the ISM Gallery Walk, showcasing the methodology, results, and impact of the Energy Saver AI project.
Key Features
- 🌱 Reduces AI energy consumption through prompt optimization
- 🔍 Analyzes prompts using T5 transformer model
- 📊 Measures and displays energy savings using Zeus library
- 💻 Web-based interface for easy accessibility
- ⚡ Real-time prompt optimization
- 📈 Statistics on token reduction and energy saved
- 🌐 Hosted on GitHub Pages for public access