Page 241 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

quicker with reliability. The decoded semantic information may either be directly consumed by
the control station, or it can be fed to a GenAI server with local knowledge base to predictively
recreate a scene corresponding to the received semantics.

If low semantic noise is ensured, then such a system can provide efficient remote visualization Change 4.2-Climate
of disaster sites.

Thus, faster accurate response planning can be ensured.

This solution will benefit better response to massive disaster thereby potentially improving
efficiency in well-being of life on land. Through better possibility of reaching out to the disaster
response agency for agile actions, this solution leads to reduced impact of aftermath of disasters.

Use Case Status: Initial PoC Phase

2�2 Benefits of use case

This solution will benefit better response to massive disaster thereby potentially improving
efficiency in well-being of life on land.

Through better possibility of reaching out to the disaster response agency for agile actions,
this solution leads to reduced impact of aftermath of disasters.

2�3 Future Work

• Our initial finding shows that the existing datasets are inefficient for the purpose, especially
considering countries like India and other developing and under-developed nations. So
we need to develop a data set.
• We plan to use LlAMA as this is an open system and tune this with the data sets for the
purpose. [3]
• Edge computing based end-to-end implementation

3 Use Case Requirements

REQ-01: It is required for the UE to have on demand access to an MEC, if it is resource
constrained.

REQ-02: It is required for the models used in transmitter-side inference, receiver side decoding
and GenAI to be seeded with similar data and training.

205