Page 190 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

2�3 Future Work

Data collection: Collect high-quality, multimodal geoscience data through open and
collaborative methods, ensuring a rich and diverse open-source dataset to support
comprehensive model training and improve GeoGPT’s effectiveness across various scenarios.

Model development: GeoGPT currently operates with parameter scales ranging from 66 billion
to 110 billion. Future initiatives aim to further optimize its architecture, create models adaptable
to diverse applications and inference devices, and open-source the model series. These efforts
will promote collaboration, drive technological innovation, and enhance accessibility to AI-
driven solutions.

Create new variations/extensions to the same use case: GeoGPT will integrate with other
ecological protection tools to provide tailored solutions for diverse environmental challenges.
Key improvements include boosting its image recognition and predictive warning capabilities,
developing microalgae classification models using microscopic images, and establishing
a spatiotemporal prediction system for red tide outbreaks. These advancements will offer
scientific support for addressing critical ecological issues.

Standards development related to the use case: Participate in ITU activities related to Natural
Disaster Management, promote the application of AI in the field of marine disaster management,
develop standards to provide norms and guidance for the industry, and facilitate the promotion
and implementation of best practices.

Setup reference tools, notebooks and simulation environment: Provide users with practical
reference tools, detailed notebooks, and simulation environments to enhance their ability to
utilize GeoGPT for recognition, monitoring, prediction, and decision-making. This will improve
the operational efficiency and usability of the platform, empowering users to make data-driven
decisions in ecological preservation efforts.

3 Use Case Requirements

REQ-01: It is required to have high-quality multimodal geoscience data for model training.

REQ-02: It is required to have scalable cloud infrastructure for AI model deployment.

REQ-03: It is required to have integration with existing marine ecosystem protection tools.

REQ-04: It is required to have real-time data processing and predictive analytics capabilities.

154