- Ultimate vision
- Start with homework but eventually it should be a multimodal learning support system that is model agnostic i. e. leverage whichever model is best available at a given time with minimal customizations at the model level
- The interactions shouldn't be limited only to voice exchanges. The system should be able to share relevant multimedia content like audio and video to help the students understand the concepts better.
- How is it different from ChatGPT?
- Personalized Learner Profiles
- Develop the learner profile over a period of time based on the student interactions
- It will be school-agnostic where the student profile carries over across institutions.
- Keep track of learning progress of the student over time in reference to an academic concept directory
- Teachers in the loop
- Teachers and other humans related to the student will have full visibility over the student's learning progress and their gaps so that the adults are also more involved in the learning process.
- Contextualize the AI content generation and interactions based on the textbook/curriculum content
- ConvoEd team will ingest the textbook content and prepare all the activities
- Teachers only have to assign with optional editing
- This will require some research and engineering effort to do with minimal human intervention
- Will we build our own AI models?
- No, we won't. We will just build the best learning infrastructure that can leverage the best in class model available. Our approach will be that of Perplexity's where the focus will be more on the user experience than the model aspects.