Inspiration
This project has a special place in our hearts. We know students struggle every day getting the adequate help for their math assignments, particularly in the topic of advanced algebra. We know it is painful when students have to wait long periods of time before their questions are answered. We believe in creating an environment implementing "live" feedback with image recognition.
What it does
This AI acts as a live tutor, integrated with text-to-speech. It uses Wolfram Alpha to solve any questions, and then compares the result to the answer provided by the student via a live recording, overlooking their paper. Having this "answer key", GUBI provides feedback in intervals about the solving process, increasing with intensity overtime.
How we built it
Using OpenAI, Mathpix, and Wolfram Alpha, we initialized "answer-keys" and ran the student response converted to latex into ChatGPT. With some prompt engineering, ChatGPT returned either "incorrect" with a hint or "correct". As progress stalled, ChatGPT returns more intensive hints. All messages are passed through OpenAI's TTS as a means to not distract the student.
Challenges we ran into
A big problem was outdated APIs, specifically WolframAlpha. Also, the inconsistency of ChatGPT proved to be a hassle.
Accomplishments that we're proud of
The ability to integrate various different tools to work harmoniously.
What we learned
Converting from idea to executable is way harder than you might think :)
What's next for GUBI.AI
Extend to all STEM principles.
Built With
- gibhub
- mathpix
- openai
- wolfram-technologies
Log in or sign up for Devpost to join the conversation.