Purpose of Blog
AI has a potentially profound impact on how we do our work. This blog elaborates on how we can get started Teaching with Geospatial AI in ArcGIS – where do we incorporate it in the curriculum, available resources, as well as possible limitations.
It is a series of 4 blogs:
How do I get started
- Educational approaches and objectives vary – please review the resources listed within this blog series to get ideas of what could be applicable to you. For those who favor books Esri Press just published “Exploring GeoAI: Tools and Workflows" (reference below) which contains step-by-step exercises. Consider using task-specific pre-trained models available in the ArcGIS Living Atlas and tutorials/courses from ArcGIS Tutorials team and Esri Academy.
- Think about how to prioritize topics or infuse existing content with AI examples – you could be faced with a scenario of “too much content; where to add AI?” in a course, so consider how existing topics/concepts could be enhanced with GeoAI approaches. An example is comparing feature extraction via traditional and deep learning methods.
- Think about prerequisites/foundational knowledge - it is important for students to understand foundational concepts before using Geospatial AI tools. This is very much course-objective dependent, but consider prerequisites ahead of time, before modifying any materials.
- Should students have basic knowledge of Remote Sensing analytics and technologies (e.g. optics, physics, platforms)?
- Should students have Programming skills?
- Should students have Statistics skills?
- The answer is normally yes to all these.
Items to keep in mind
- Use a smaller dataset as proof of concept – test models on small areas then extrapolate to what it would take to analyze larger areas. There are several factors contributing to the successful use of pre-trained models. Firstly, the data used is crucial; it is best to use data that is similar in type, resolution, and location as the training data.
- Be prepared for multiple attempts- one pass may not give satisfactory results; it usually takes more than one try. Improvements can be made through techniques such as manual editing, retraining the model, transfer learning or experimenting with other deep learning models.
- Ensure students understand limitations of using AI in learning - importance of developing a healthy critical view of data and outputs produced with AI tools. The results may be faster but are they better or even correct?
Governance, Compliance, Ethics
Check your institutional policies and resources as there is considerable variation and guidance on using AI.
- Check your institutional AI resources (if any exist) – some institutions offer guidance such as the below.
- AI Guidance, Approved AI Tools, and AI Training
- AI project checklist to gather information about AI projects
- AI-related external workshops and webinar links
- AI Learning Community space to share, learn, and discuss teaching with AI lessons learned, challenges, and solutions.
- Provide action items and safeguards for students (example items below).
- Sample Syllabus disclaimers, outlining user responsibility, ethical use, data security and privacy, potential risks, support. Again, those vary institution by institution.
- IT Risks Assessment Processes (checklists, procedures) – also vary by institution.
- Department or institution level guidelines.
- Some institutions employ policies that everyone must undergo, or undergo if certain conditions are met, such as access to institutions systems, logins/authentication, physical hosting of data on institutional IT systems, etc.)
Technology/Hardware
For certain processes/workflows, consider necessary hardware requirements:
- GPU requirements should be considered (Are GIS lab computers up-to-date?)
- Hardware access should be thought of before deploying in class (i.e. would students do work in computer labs, on their own computers, etc.)
- Additional libraries for performing deep learning in ArcGIS Pro need to be installed before using the tools.
- ArcGIS Online also supports GeoAI however network bottlenecks should be considered if, say, 20 students are accessing the same model.
Resources, where to go for help
Feel free to refer to other blogs in the series:
Key points when thinking about the next steps:
- Do not be afraid – it is easy to get started, even in introductory courses.
- The tools are accessible, already within ArcGIS.
- Think about how and whether AI is a good choice for what is being taught.
Please take a look at the below resources, which could be helpful in the classroom, providing students with an easier learning experience. Feel free to comment on what else could be helpful.
- Getting started
- Learn Tutorials and Esri Academy (search for AI, GeoAI, AI Assistants, example tutorial series)
- Documentation and sample packages
If you need to speak to a person, please reach out:
- United States
- Outside United States