AI-ML

<centre>

The CSDMS AI/ML Initiative

Objectives

The Artificial Intelligence / Machine Learning Initiative aims to help with the application of these technologies to the science of Earth Surface Dynamics Modeling.

Four main activities are identified:

1. Disseminate knowledge on techniques and applications. Especially to transmit know-how between users with different levels of experience and training, and from one domain of earth surface science to another. The aim is to make uptake of ML more efficient/time-effective for new entrants, and then extend the sophistication of applications to Earth Surface Dynamics Modeling (ESDM).
2. Foster the community using the email list, workshops, newsletter, tweets, generating proposals, (?)Github worked examples, etc.
3. Create an ideas/examples data-space on CSDMS pages for feature layers, preferably continent- or world-wide. This is probably the single most useful component for the membership because conceiving and making suitable training/testing feature layers for ML is a creative process.

Contacts

The Initiative is part of the Cyberinformatics CSDMS Group. The Lead scientists will be Jeff Obelcz (NRL Stennis), Chris Jenkins (INSTAAR, CU Boulder), Kelly Kochanski (CU Boulder), Evan Goldstein (UNC Chapel Hill).

Members

• Members

Activities

1. At May 2019 Annual Meeting of CSDMS - two keynote talks on ML projects, and a plenary panel meeting on the topic, lead by researchers with publications in the field.
2. A Conversation Portal using Twitter and Google Groups - releasing in March 2019 so members can interact, learn, and practice with ML tools.
3. Gathering and implementing advice from members on what they need, what CSDMS can help with, to further the technology in the modelling of Earth Surface Dynamics

Background

A beginning to the activities took place on 24 May 2018 in the CSDMS Annual Meeting with a forum to decide whether to create an Initiative Group and then on the courses of action. Creation of the group was voted for overwhelmingly. The forum was convened by Chris Jenkins and Jeff Obelcz and attended by over 40 people.

Title: "With Artificial Intelligence & Machine Learning - What Lies Ahead for Earth Surface Modeling ?"

The proceedings are available online through CSDMS: Wiki, Agenda and Notes. Following that the CSDMS Executive decided to form an Initiative Group as a matter of urgency within the Cyber Working Group, but which may be rearranged in the CSDMS structure at a future time.

Four main questions were put to meeting:

1. What can AI and ML currently do that might benefit Earth Surface Dynamics Modeling ?
2. What is the relationship between Process Modelling and AI/ML ?
3. How should CSDMS Community Respond to the Appearance of AI/ML
4. What Earth Surface Dynamics Modeling-related tasks are not suited for AI/ML? Why?