A couple of additional ideas that I think could be useful to the community: Deep learning in general and applications for data assimilation (or data fusion). For example, a hackathon where the focus is on using sparse in-situ observations and combining them with large-scale models. This is a problem common across all communities: ocean, land, and atmosphere.
A focus on foundation models (which are also deep learning), and a hackathon on adapting them for downstream tasks (I’m not familiar with foundation models for the ocean, but I’m sure there will be a few).
My personal opinion: don’t make it specific to GenAI; better to keep it open to deep learning more broadly, as GenAI fits naturally within foundation models and can also be relevant to data assimilation workflows. Developing GenAI from scratch would probably not be practical for a hackathon, but adapting existing models, e.g. foundation models or other deep learning models, would be much more feasible.
Perhaps I was too specific with “Generative AI”. I mean using all these LLMs that are around to help with coding.
My worry tho is that most of them require payed subscriptions to be used with VS Code or other tools so I’m not sure how it will happen. But perhaps I’m missing something?
Oh, I see. My brain was thinking about GenAI for science, not LLMs, I meant developing generative diffusion models or Generative Adversarial Networks (GANs). Having a session where people share their experiences with different AI tools for research, coding and writing.. would be greate to include at every opportunity
