Course material (2024 edition)
Target level: BSc 3rd year (BSc Cognitive Science and AI)
Taught in the Department of Cognitive Science and AI at Tilburg University.
This course will provide the skills and knowledge to apply AI and data science in multiple ways to help nature and the environment. The biodiversity crisis and climate crisis are complex and interconnected: luckily there are many ways that technology can help to monitor the natural world, and to help society have a more positive impact. After successful completion of this course, students will be able to:
- Explain multiple current/novel ways in which AI can be used for biodiversity/ecology/environment.
- Critically evaluate potential data-driven interventions in natural environments, for their benefits and impacts.
- Describe good practice and common pitfalls in machine learning and data processing, within the realm of biodiversity/ecology/environmental data.
- Implement algorithmic data analysis for biodiversity/ecology/environment in Python.
- Analyse the performance of AI algorithms on datasets relating to the natural world.
This course focusses on a diverse set of applications of tech for nature, in each case studying how data science and AI methodologies can be used. We also encourage a critical and comparative approach, by looking at the impacts as well as the benefits of tech for nature, and considering machine learning good practices. The course assumes some familiarity with programming (Python) and with AI concepts, and explores the topics through computer-based data/AI practical work.
- Intro: biodiversity, climate, data science, AI
- Deep learning for images of wildlife
- Seeing from above: Remote sensing
- Deep learning for audio
- Citizen science
- Electricity generation/grids. And: Cost-benefit & critiques -- Lecture 6 slides 2023 -- Lecture 6 video 2023
- Devices in the wild
--mid-term-- - Tracking movement: Biologging
- Bad tech, versus nature-based solutions -- Lecture 9 slides 2023 -- Lecture 9 video 2023
- Advanced topics 1
- Advanced topics 2
- Synoptic, exam, Q&A
To join this course you must already have:
- Experience in python programming
- Knowledge of machine learning, e.g. through CSAI courses on ML/DL
(a) 60% final exam
(b) 30% individual coding project
(c) 10% mid-term "paper review", in two parts: 5% is individual paper review video; 5% is a group paper report
- "Perspectives in machine learning for wildlife conservation", Tuia et al. 2022.
- "Emerging technologies revolutionise insect ecology and monitoring", van Klink et al. 2022.
- "Computational bioacoustics with deep learning: a review and roadmap", Stowell 2022.
- "Tackling Climate Change with Machine Learning", Rolnick et al 2019.
Let me address one thing: "AI... for Nature and Environment"? Really? Does nature really need more AI? ... Well in a way, no. AI is not the solution. Land management, good politics, quitting oil, and quitting beef -- they're all much more important. However -- AI and modern data technologies are crucial to effective management of almost all the good solutions, even "nature-based solutions". In this course we focus on the benefits as well as the costs, and various different ways in which machine learning and related methods might help. This course is intended for everyone out there who has developed some skills with data science and machine learning, and wants to use them for good. For example, if you're thinking about how you can move into a new career in which those skills are actually helping with some of the world's biggest problems: climate and biodiversity.
I highly recommend both Climate Change AI and WildLabs which organise online events, courses, links to interesting stuff, and more.
This material is copyright (c) Dan Stowell 2023, and is published under the Creative Commons CC-BY-NC-SA 4.0 license.