workshop summary
Causality and efficient data exploitation
Core Inquiry: Through a set of talks by domain experts from academia and industry, the workshop highlighted the need for better exploitation of the available multi-modal weather, health and climate data. The crucial question of what are the appropriate hardware resources needed to enable AI to exploit this data, balancing the requirements for powerful machines and environmental concerns, was raised.
Scientific Principles and AI: The role of AI in finding not just correlations, but causality between data collected and target climate, weather or health events was discussed. The workshop also highlighted the need for AI to be scalable, and make sense of the complex relationships between the different elements of the systems studied.
Vision: The vision set forth was that of a robust, scalable, comprehensive and explainable AI model for the Earth, capable of short and long term predictions of regular and outlier events, enabling a better understanding of the mechanisms behind weather and climate as well as early action in the case of disasters.
Actionable insights and future directions
In a consensus between the AI and domain experts, a first step in achieving the ambitious vision set forth requires the development of AI models able to find causal relationships in the constently evolving data landscape available. It was aknowledged that research funding into specialised, and potentially novel, hardware will be needed, and it should go hand in hand with funding into traditional, domain specific research and equipment. A strong collaboration between AI and domain experts, was highlighted as paramount to ensure that data is utilised efficiently and to set benchmarks enabling the validation of the AI models developed.