Research Questions (fine-tuning for desktop apps / training from tutorial videos / sandbox) #1
Description
Dear research group,
thank you for the awesome paper you just published! 🎉
For our educational live linux distro, we have been looking to implement an AI desktop assistant that can support a novice solving practical computing tasks and guide young learners towards acquiring solid machine interaction skills. Since the release of cogagent last year we've been keeping our eyes open and now your model is very to close to what could enable such assistive functionality.. so here are two three questions:
- Can your model be fine-tuned for specific desktop software (like inkscape, freecad, libreoffice, firefox …) ?
- Could it be made to train on tutorial videos?
- What about a sandbox for the agent to explore the application by itself and try to complete tutorials?
best regards