# EGU 2025 Townhall SWUNG

## 2025-04-17

* Switch from panel-based discussion to more interactive
* Use Mentimeter or some interactive tool to promote discussion

- Open with SWUNG slides (15 minutes )
- Introducing organization - goals and motivation
- Find map that Matt generated in one of the transforms with location of participants. Where it might be?
- Google Drive?
- Ask Matt?
- Here: https://github.com/softwareunderground/swung-org/blob/main/agm/2022-04-28.pdf
- Mattermost statistics?
- Easy to get from the last month.
- Emphasis in Transforms
- History of Transforms
- number of participants / tutorials output
- where to find the existing material

- Mentimeter ideas :
- who has heard of SWUNG before
- how often do you work with open-source software
- 1-5 slider ("never" to " all the time")
- do you use LLM in your daily task
- 1-5 slider ("never" to "all the time")
- How do you use it ?
- multiple choice (other?)

- What type of environmental impacts are you aware of (multiple choice)
- Carbon emission / greenhouse gases
- Energy consumption
- Water consumption
- Other: ....

- In the context of learning how to program:

- Slide with some references

- Split ppl in groups if we have > ~20 ppl

Big questions:
- Do LLMs pose a threat to the global efforts to mitigate climate change?
- Do LLMs provide a viable way of learning computer science and software development skills?
- Is the learning process becoming more isolated through usage of LLMs?
- What ethical uses of LLMs and AI do exist?
- Can communities like Software Underground overcome the difficulties that LLMs have by fostering human-human interactions?

Last thing : short setences

## 2025-03-26

* Ask EGU if we can invite panelists that are not registered for EGU. Online participation?
* Order Swung swag earlier than late. Stickers take ~10 days to be shipped!
* The next SWUNG board meeting clashes with EGU. We should keep that in mind.

## 2025-03-19

* Slides (< 10 minutes )
* What's swung?
* Start interactivity
* How it started (grassroots)
* **Our goals and motivation**
* How much support
* Free tutorials in Transform
* Projects involved
* Visible impacts (quotes, anecdotes and positive outcomes)
* Opening the discussion :
* One big question
* Bibliography analysis of existing papers that talk about it. (visualize the hype)
* Quantity vs Quality
* Introduce environmental impact (choose 1-2 articles that we can showcase)
* Talking points, trigger questions

* Format
* Panel discussion
* 3 panelists
* 20 mins for panelists only, moderated by questions from the moderators (20 minutes)
* Panelists?
* Paul Cleverley: https://paulhcleverley.com/about/
* Wim Vanderbauwhede: https://wimvanderbauwhede.codeberg.page/
* How to make it different from the [EGU panel](https://meetingorganizer.copernicus.org/EGU25/session/54301)?
* Focus on open-source software development and community building?

* Discussion topics:
* Usage of AI for software development in geosciences
* Do the LLMs really understand open-source code and its implementation?
* Environmental impact
* What if LLMs would inform users about climate impact?
* Communities as an alternative
* Collective building of knowledge

* Audience discussion ( ~30 minutes )
* Interactive questions to the audience
* Backup questions are a must
* Ask the audience how do they use LLMs and AI.

Closing statements
- invite SWUNG members, keep in touch, participate.

### TODO

* [ ] approval swung board
* [ ] stickers!

- what do we do as a community ? "Promote concious use"
* What is a _concious use_?
* Realistic expectations. What are viable outcomes from this?
* These could be the outro questions.
-

Pros:
- Different prespectives
- Lifts the weight of work for moderators

One topic that will differentiate us from the existing panel discussion is introduce the environmental impacts. For this, we can invite a specialist (Wim Vanderbauwhede)

---

# Proposal

## Requirements

- [x] Title
- [x] Short description (300-400 words) outlining:
- motivation
- target audience and
- expected outcomes of meeting.
- [x] Which Programme Groups is the meeting related to?
- EOS, ESSI, ERE
- [x] Keywords (2 to 5)
- [x] Conveners:
- Santi
- Andrea
- More ppl are welcome to join!
- [x] Preferred day (Mon to Fri)
- **Tue**
- [x] Estimated number of attendees:
- [x] 0–100
- [ ] 101–219
- [x] Do you anticipate ordering catering for your Townhall meeting?
Note: catering must be organized by you, at your own expense
- Nope.. sorry.

## Title

Software Underground Meetup: Ethical use of LLMs in scientific research and open-source software development

## Keywords

* `open-source software`
* `artificial intelligence`
* `climate change`
* `Scientific community`
* `solid Earth`

## Short description

This first-ever Software Underground Townhall Meeting invites earth scientists
interested in open-source software who are concerned about usage of Large
Language Models (LLMs) in learning processes, software development and its
environmental impacts.
We will discuss how communities like Software Underground can be an alternative
resource by creating collaborative environments and strengthening professional
relationships as well as promoting reliable and reproducible research
practices.

Software Underground is a grassroots community of digital subsurface
professionals engaged in a diverse array of topics, united by a common interest
in open-source scientific software and open science.
Our mission is to offer technical and professional development opportunities to
the global earth sciences community while advocating for the use of open-source
digital technologies in solid earth investigations.

LLMs are presented as time-saving tools for writing tasks and code generation.
However, one of the pushing concerns regarding the wide use of LLMs is their
environmental impact: from their increased demand of energy and clean water, up
to the predicted amount of e-waste they'll generate in future years.
The other concerning factor is their ability to produce inaccurate or even
false statements, and how they replicate existing biases in our society.
This is particularly relevant when learners use them to acquire new knowledge.
They could also make the learning process more isolated, hindering the
interaction with peers to challenge acquired knowledge or to build new
knowledge in a collaborative environment.

During this meeting we'll discuss about the ethical use of LLMs and other AI
tools in scientific research and software development.
We'll explore the benefits of these novel tools can bring to scientific
research, and what type concerns they raise.
We'll also discuss how communities of peers can create alternative paths to
gather new knowledge, make research more collaborative and build stronger
professional relationships.

The discussion will be encouraged through a set of questions like:

- Do LLMs pose a threat to the global efforts to mitigate climate change?
- Do LLMs provide a viable way of learning computer science and software development skills?
- Is the learning process becoming more isolated through usage of LLMs?
- What ethical uses of LLMs and AI do exist?
- Can communities like Software Underground overcome the difficulties that LLMs have by fostering human-human interactions?

The conveners will promote group dialogues designed to encourage meaningful
conversation between all participants, which will be captured in collaborative
shared notes.

## References

* ['Thirsty' ChatGPT uses four times more water than previously
thought](https://www.thetimes.com/uk/technology-uk/article/thirsty-chatgpt-uses-four-times-more-water-than-previously-thought-bc0pqswdr)
* [Vanderbauwhede (2024). Estimating the Increase in Emissions caused by
AI-augmented Search](https://doi.org/10.48550/arXiv.2407.16894)
* [Vanderbauwhede (2023). Frugal Computing -- On the need for low-carbon
and sustainable computing and the path towards zero-carbon computing](https://doi.org/10.48550/arXiv.2303.06642)
* [Wang et al. (2024). E-waste challenges of generative artificial
intelligence](https://doi.org/10.1038/s43588-024-00712-6)
* [Luccioni et al. (2024). Power Hungry Processing: Watts Driving the Cost of
AI Deployment?](https://doi.org/10.48550/arXiv.2311.16863)
* [Lehmann et al. (2024). AI Meets the Classroom: When Does ChatGPT Harm Learning?](https://doi.org/10.48550/arXiv.2409.09047)
* [](https://greenscreen.network/en/blog/within-bounds-limiting-ai-environmental-impact/)
* [Lee et al. (2025)](https://www.microsoft.com/en-us/research/wp-content/uploads/2025/01/lee_2025_ai_critical_thinking_survey.pdf)
---

## Draft ideas

Split the conversation in two, so we don't talk for an hour of the same topic.
* AI and its environmental and ethical issues
* Communities as an alternative to overcome these drawbacks

Target audience:
* earth scientists interested in open-source scientific software for solid-earth investigations, open-science and reproducible research.
* concerned about usage of AI in learning, software development and its environmental impacts.

Swung motivations?
* advocate for the use of open-source digital technologies in solid-earth investigations.

What will we do?
* Discussion of ethical use of AI in scientific research and development of scientific open-source software.

Environmental issues:
* Increasing global energy consumption in a world with climate crisis.
* Requires large amounts of clean water.
* Future generation of e-waste.

Challenges of AI as a learning tool:

* Biased results. Correctness of answers is not guaranteed. Challenges of unsupervised learning with AI: can lead to incorrect "knowledge".
* "Acquiring inaccurate knowledge"
* Isolation of the learners. The learning process happens in isolation, without building bonds with colleagues.

What is AI useful for?

* Would be nice to have some useful cases for AI. At least state the question about ethical uses of AI.
* AI like segmentations are useful for scientific applications.

We should focus the conversation on LLMs and not AI in general.

> Put this somewhere: (As we often meet online, our motivation is to meet in person and also broaden our community's outreach to those attending EGU.)

> The session will facilitate discussions that will be captured in collaborative shared notes and group dialogue designed to encourage meaningful conversations.