From 02b384f66fa873b1f66c0b328212fa76a5beaf8c Mon Sep 17 00:00:00 2001 From: cmohge1 Date: Thu, 18 Apr 2024 14:12:14 +0000 Subject: [PATCH] deploy: ca97442f8e97a0a58dd8c790cd0f670e39840ffd --- 404.html | 2 +- index.html | 8 ++++---- sitemap.xml | 2 +- toolkit.html | 2 +- toolkit/advocating-within-your-institution.html | 6 +++--- toolkit/case-studies.html | 4 ++-- toolkit/climate-change-faqs.html | 2 +- toolkit/decision-trees.html | 4 ++-- toolkit/grant-writing.html | 2 +- toolkit/introduction.html | 6 +++--- toolkit/maximal-computing.html | 8 ++++---- toolkit/minimal-computing.html | 4 ++-- toolkit/teaching.html | 2 +- toolkit/working-practices.html | 4 ++-- 14 files changed, 28 insertions(+), 28 deletions(-) diff --git a/404.html b/404.html index bdd7a61..d3f8278 100644 --- a/404.html +++ b/404.html @@ -1,2 +1,2 @@ -404 Page not found | DHCC +404 Page not found | DHCC

404

Page Not Found

DHCC

\ No newline at end of file diff --git a/index.html b/index.html index e313c0a..7e4363f 100644 --- a/index.html +++ b/index.html @@ -1,4 +1,4 @@ -Introduction | DHCC +Introduction | DHCC
Introduction @@ -9,7 +9,7 @@ #

Digital Humanities and the Climate Crisis: A Manifesto #

Digital Humanities and the Climate Crisis: A Manifesto. “The digital is material. As digital humanists, every project we create, every software application we use, every piece of hardware we purchase impacts our environment. […] As humanities researchers, it is also our role to probe the values, the power structures, and the future imaginaries that underpin sustainable solutions.”

UK-Ireland Digital Humanities Association #

The Digital Humanities Climate Coalition is a Community Interest Group of the UK-Ireland DH Association. “The Association builds on existing research conducted by the partner institutions and brings together different stakeholders to consider and interrogate critically the concepts of sustainability, inclusivity, training, advocacy and career progression, among other key questions. Institutions and individuals engaged with the Association, across the UK and Ireland, will work to further develop a collaborative vision for the field, and create new and sustainable long-term partnerships with the international community.”

Workshops and Talks -#

Members of the DHCC have given talks and workshops related to this toolkit in various contexts including DH Benelux, the British Library, National Library of Scotland, the Canadian Health Libraries Association, and the National Archives. If your organisation is interested in finding out more, contact one of the coordinators, Jo Lindsay Walton, James Baker, Christopher Ohge, or Lisa Otty.

A Researcher Guide to Writing a Climate Justice Oriented Data Management Plan -#

This guide aims to help researchers think through the climate-related implications of decisions about data management and project planning. It includes an annotated DMP.

DHCC Information, Measurement and Practice Action Group. (2022). A Researcher Guide to Writing a Climate Justice Oriented Data Management Plan (v0.6). Zenodo. button

Beekeeper

See also Grant Writing.

Calendar -February 25, 2024
Edit +#

Members of the DHCC have given talks and workshops related to this toolkit in various contexts including DH Benelux, the Norwegian Technical University, the British Library, National Library of Scotland, the Canadian Health Libraries Association, and the National Archives. If your organisation is interested in finding out more, contact one of the coordinators, Jo Lindsay Walton, James Baker, Christopher Ohge, or Lisa Otty.

A Researcher Guide to Writing a Climate Justice Oriented Data Management Plan +#

This guide aims to help researchers think through the climate-related implications of decisions about data management and project planning. It includes an annotated DMP.

DHCC Information, Measurement and Practice Action Group. (2022). A Researcher Guide to Writing a Climate Justice Oriented Data Management Plan (v0.6). Zenodo. button

Beekeeper

See also Grant Writing.

\ No newline at end of file diff --git a/sitemap.xml b/sitemap.xml index 06174e2..0be871c 100644 --- a/sitemap.xml +++ b/sitemap.xml @@ -1 +1 @@ -/toolkit/introduction.html2024-01-30T13:25:29+00:00/toolkit/decision-trees.html2024-02-29T17:35:34+00:00/toolkit/minimal-computing.html2024-02-29T17:32:24+00:00/toolkit/maximal-computing.html2024-03-28T19:43:02+00:00/toolkit/grant-writing.html2023-12-06T15:59:22+00:00/toolkit/working-practices.html2024-03-08T11:40:32+00:00/toolkit/advocating-within-your-institution.html2024-03-02T19:15:12+00:00/toolkit/climate-change-faqs.html2024-02-26T11:28:14+00:00/toolkit/teaching.html2024-03-02T15:52:01+00:00/toolkit/case-studies.html2024-02-22T16:23:17+00:00/toolkit.html2024-03-28T19:43:02+00:00/2024-02-25T16:08:39+00:00 \ No newline at end of file +/toolkit/introduction.html2024-01-30T13:25:29+00:00/toolkit/decision-trees.html2024-02-29T17:35:34+00:00/toolkit/minimal-computing.html2024-02-29T17:32:24+00:00/toolkit/maximal-computing.html2024-03-28T19:57:34+00:00/toolkit/grant-writing.html2023-12-06T15:59:22+00:00/toolkit/working-practices.html2024-03-08T11:40:32+00:00/toolkit/advocating-within-your-institution.html2024-03-02T19:15:12+00:00/toolkit/climate-change-faqs.html2024-02-26T11:28:14+00:00/toolkit/teaching.html2024-03-02T15:52:01+00:00/toolkit/case-studies.html2024-02-22T16:23:17+00:00/toolkit.html2024-03-28T19:57:34+00:00/2024-04-18T15:11:47+01:00 \ No newline at end of file diff --git a/toolkit.html b/toolkit.html index 6afce53..c94f975 100644 --- a/toolkit.html +++ b/toolkit.html @@ -1,4 +1,4 @@ -Toolkits | DHCC +Toolkits | DHCC
Toolkits diff --git a/toolkit/advocating-within-your-institution.html b/toolkit/advocating-within-your-institution.html index d61289a..938c12c 100644 --- a/toolkit/advocating-within-your-institution.html +++ b/toolkit/advocating-within-your-institution.html @@ -1,10 +1,10 @@ Advocating within your Institution | DHCC +— Roy Rappaport, quoted in Steward Brand’s Cybernetic Frontiers (1974) Put your institution in context # Our institutions are taking action. Most European universities have now set net zero targets and milestones, as part of broader sustainability strategies. They are integrating climate risks and opportunities into governance and strategy, and embedding climate change across their curriculums.">Advocating within your Institution | DHCC
Advocating within your Institution diff --git a/toolkit/case-studies.html b/toolkit/case-studies.html index 68fc0c5..49c0a44 100644 --- a/toolkit/case-studies.html +++ b/toolkit/case-studies.html @@ -1,6 +1,6 @@ Case Studies | DHCC +Highlights # GLAM sector organisations face many challenges in decarbonisation, including need for financial support to decarbonise, and expertise in carbon accounting and management planning. Carbon accounting and management support needs to be offered accessibly and at scale, and there is a role for automated approaches to help this happen.">Case Studies | DHCC
Case Studies diff --git a/toolkit/climate-change-faqs.html b/toolkit/climate-change-faqs.html index d22505d..f651e36 100644 --- a/toolkit/climate-change-faqs.html +++ b/toolkit/climate-change-faqs.html @@ -1,4 +1,4 @@ -Climate Change FAQs | DHCC +Climate Change FAQs | DHCC
Climate Change FAQs diff --git a/toolkit/decision-trees.html b/toolkit/decision-trees.html index ffe152e..59762c1 100644 --- a/toolkit/decision-trees.html +++ b/toolkit/decision-trees.html @@ -1,6 +1,6 @@ I Want To ... | DHCC +Our first step below is to build a list of needs, and to signpost relevant evidence and recommendations within the toolkit.">I Want To ... | DHCC
I Want To ... diff --git a/toolkit/grant-writing.html b/toolkit/grant-writing.html index fac8891..030355d 100644 --- a/toolkit/grant-writing.html +++ b/toolkit/grant-writing.html @@ -1,4 +1,4 @@ -Grant Writing | DHCC +Grant Writing | DHCC
Grant Writing diff --git a/toolkit/introduction.html b/toolkit/introduction.html index 35e9945..4e1bd54 100644 --- a/toolkit/introduction.html +++ b/toolkit/introduction.html @@ -1,8 +1,8 @@ Introduction | DHCC +To meet climate targets, the world needs to cut net carbon emissions by about half by 2030, and achieve net zero carbon by 2050.">Introduction | DHCC
Introduction diff --git a/toolkit/maximal-computing.html b/toolkit/maximal-computing.html index a0717cc..c641199 100644 --- a/toolkit/maximal-computing.html +++ b/toolkit/maximal-computing.html @@ -1,6 +1,6 @@ Maximal Computing | DHCC +Key Recommendations # Develop a clear understanding of when it is appropriate to use AI and blockchain for a humanities research problem. When you do need to use maximal computing, consider using less-intensive processes like using smaller models and adapt existing infrastructures. Understand how to use efficiency metrics like T-scores, confusion matrices, accuracy, recall, Area under the ROC curve, LOGLOSS.">Maximal Computing | DHCC
Maximal Computing @@ -68,7 +68,7 @@ criterion alongside accuracy” is what Schwartz et al. call Green AI.

Contrasting with much of the evidence assembled here, this 2022 preprint offers an optimistic assessment of future model training carbon impacts, and proposes some best practices (selecting sparser models, using optimised processors such as TPUs and recent GPUs, using hyperscalers rather than on-prem data centres, and choosing data centres powered by the cleanest energy).

Understanding Cloud emissions #

It also makes a big difference where (and when) you train and deploy your model. Climatiq offers a series of introductory articles to Cloud emissions:

Datacentre sustainability is strongly influenced by location, and the local availability of green electricity. Electricity Maps is an open source visualization project to help understand the climate impacts of energy use.

The Science Based Targets initiative sectoral guidance for ICT includes a focus on datacentres. To get a rough understanding of where the Cloud’s emissions fit into ICT emissions as a whole, the order goes (from largest impact to smallest): manufacturing user devices, running user devices, running data networks, running data centres, manufacturing data networks, manufacturing data centres. This doesn’t include crypto though. (We’d like to include some more granular recent data here, if you have any to add).

The environmental pledges and policies of the cloud giants (Microsoft, Amazon, and Google/Alphabet) deserve scrutiny. At the time of writing, the differences between them are significant, but none deserves enthusiastic praise.

Learning about Cloud emissions will quite quickly take you into the realm of energy generation and the power grid. For example, the 24/7 carbon-free energy approach, supported by a UN compact, appears to be a big improvement on previous highly financialised ‘annual matching’ approaches. The grid-aware computing approach also seems to improve substantially on carbon-aware computing.

FinOps and GreenOps #

Within business, there is already a considerable body of theory and practice around cloud optimisation, associated with the FinOps approach. FinOps actions taken to reduce cloud spend (e.g. reservation management, data lifecycle management, continuous rightsizing) often have the unintended consequence of reducing carbon emissions too (although that’s not always the case).

Interestingly, FinOps is also not just about cost in a narrow sense, but about multidimensional value (for example, the “Iron Triangle” of fast, cheap, and good service). FinOps practitioners are interested in the various trade-offs and co-benefits of using the Cloud in different ways.

This means there is good potential to integrate sustainability (and even climate justice) into FinOps practice. By bringing together financial, IT, and sustainability teams under a common objective, a FinOps Centre of Sustainable Excellence, or a GreenOps Centre of Excellence, might allow these traditionally siloed disciplines to collaborate and share expertise.

However, the GreenOps approach is still emerging. Sustainability is not yet a prominent theme in FinOps literature. As you might expect, FinOps also leans techno-solutionist, so there is potential for tensions with perspectives such as digital sobriety, Minimal Computing, or degrowth.

Major cloud providers offer tools that anticipate a FinOps perspective, including AWS Trusted Advisor, Microsoft Azure Cost Management, GCP Cost Management.

Other terms include DevGreenOps (Greenspector) and DevSustainOps (Eric Zie, Decarbonise the Digital).

Further reading:

Quantum computing +(2023).
  • Greenspector, How do you go from DevOps to DevGreenOps?
  • GreenPixie, What is GreenOps?
  • Greenly, FinOps and GreenOps, How Do They Relate?
  • The New Stack, What is GreenOps? Putting a Sustainable Focus on FinOps (2023)

    • Quantum computing #

    This isn’t something we have looked into yet, but we would appreciate any evidence or resources. A cursory exploration of the intersection of quantum computing and climate suggests quite a lot of positive framings, although also quite clearly some significant hype. Here’s one interesting extended blog post:

    Pinto, Karan, ‘Quantum Tech for Better Climate: A Thesis’, Entangle Quantum (2022)

    Mini case study #

    GPT-n workshops case study #

    In some recent workshops to introduce participants to GPT-n text generators, participants fine-tuned a pre-trained model using texts of their choosing.

    There were good pedagogic reasons to let everyone choose their own input text, e.g. from Gutenberg, for the fine tuning: it turns the activity into a more exciting experiment. It also made sense for me to do the activity myself, because (1) it's easy to forget details when you're describing something from memory and notes, and (2) I know from experience that certain fine-tuning texts will "work well," so we would have at least one interesting output to discuss at the end.

    But in the short time available, participants often chose arbitrary and / or "obvious" input texts (e.g. Shakespeare, or an author they were primed to choose because they had been mentioned earlier in the workshop). Likewise, pushed for prep time, the workshop leader realised they were tending to use the same input texts again and again. Furthermore, we would generate hundreds of pages of text, and then only browse the first few - and never use the rest for anything.

    My new rule of thumb: if I am running a computationally intensive process for educational or demo purposes, do so in a way that doubles as research (in a loose sense of "research").

    For the latest iteration of the workshop we have:

    • Explicitly included discussion of carbon costs.
    • Used a recording of myself doing the fine-tuning and text generation.
    • Created a shared folder of potential fine-tuning inputs which I am interested in for various reasons - participants could choose one of these or pick their own.
    • Built in time to discuss actual immediate use cases of text generation, and what questions we might ask of the output (and how many pages we needed to generate).

    In some ways the workshop felt clunkier, and I had to cut some content which I liked. But it still felt worth it to include very basic sustainability considerations.

    If I run similar workshops in the future I hope to:

    • Find out more about the actual carbon footprint of these processes.
    • Explore a format where small groups collaborate, running just one fine-tuning per group.
    • Explore choosing from a variety of already fine-tuned models, rather than running new fine-tunings.
    • Seek out others who may already want synthetic texts based on specific fine-tuning inputs, and offer to do this for them as part of the workshop.
    • Continue to think about proportionality, and if the carbon cost does seem too high, then replace the workshop with something else entirely.

    I don't think that the energy savings will be extensive. When I think of all the users of art AIs, generating thousands and thousands of images just out of curiosity or a playful compulsion, it feels like a drop in the ocean. But it also feels important to model responsible behaviour. A minimal set of principles for using AI to generate text or art might involve:

    • Spend the time to articulate what I am trying to do and why (not just “to try it out” or “to see what happens”).
    • Whenever possible, combine different purposes (try out an experimental approach in a way that may also contribute to a particular project, etc.).
    • Whenever possible, work in an open and shared way, so that others can benefit from my use of the AI.

    Adapting existing infrastructures @@ -97,6 +97,6 @@ pipelines being deployed. We need to raise awareness of the impacts of continually triggering what might be a resource intensive process.

    Further reading #

    Blair, Gordon S. ‘A Tale of Two Cities: Reflections on Digital Technology and the Natural Environment’. Patterns 1, no. 5 (14 August 2020). [https://doi.org/10.1016/j.patter.2020.100068].

    Kazansky, B., Karak, M., Perosa, T., Tsui, Q., Baker, S., and The Engine Room. (2022). At the confluence of digital rights and climate & environmental justice: A landscape review. Available at: [https://engn.it/climatejusticedigitalrights]

    Knowles, Bran, Kelly Widdicks, Gordon Blair, Mike Berners-Lee, and Adrian Friday. ‘Our House Is On Fire:The Climate Emergency and Computing’s Responsibility’. Communications of the ACM, 2 December 2021. https://eprints.lancs.ac.uk/id/eprint/162995/.

    Schwartz, Roy, Jesse Dodge, Noah A. Smith, and Oren Etzioni. ‘Green AI’. Communications of the ACM 63, no. 12 (17 November 2020): 54–63.

    Shervin Minaee, 20 Popular Machine Learning -Metrics. Towards Data Science, 28 Oct 2019. [Accessed 2022-04-06].

    Shao, Xiaotong, Zhongbin Zhang, Ping Song, Yanzhen Feng, and Xiaolin Wang. 2022. ‘A Review of Energy Efficiency Evaluation Metrics for Data Centers’. Energy and Buildings 271 (September): 112308. https://doi.org/10.1016/j.enbuild.2022.112308.

    Strubell, Emma, Ananya Ganesh, and Andrew McCallum. 2019. ‘Energy and Policy Considerations for Deep Learning in NLP’. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 3645-50. Florence, Italy: Association for Computational Linguistics. https://doi.org/10.18653/v1/P19-1355.

    Tremayne-Pitter, Daniel (dir.) (2022). Clouded: Uncovering the Culture of the Cloud Documentary film.

    Xu, Jingjing, Wangchunshu Zhou, Zhiyi Fu, Hao Zhou, and Lei Li. 2021. ‘A Survey on Green Deep Learning’. https://doi.org/10.48550/ARXIV.2111.05193.

    \ No newline at end of file diff --git a/toolkit/minimal-computing.html b/toolkit/minimal-computing.html index 42d3803..eabc1e1 100644 --- a/toolkit/minimal-computing.html +++ b/toolkit/minimal-computing.html @@ -1,8 +1,8 @@ Minimal Computing | DHCC +This section introduces you to minimal computing principles. Minimal computing is a set of principles and practices that aim to reduce both environmental impact and barriers to access and engagement. It offers an important set of thinking tools to make responsible, frugal, and nuanced digital decisions.">Minimal Computing | DHCC
    Minimal Computing diff --git a/toolkit/teaching.html b/toolkit/teaching.html index 0494d41..2ef632d 100644 --- a/toolkit/teaching.html +++ b/toolkit/teaching.html @@ -1,4 +1,4 @@ -Teaching | DHCC +Teaching | DHCC
    Teaching diff --git a/toolkit/working-practices.html b/toolkit/working-practices.html index e792d1f..eac9764 100644 --- a/toolkit/working-practices.html +++ b/toolkit/working-practices.html @@ -1,6 +1,6 @@ Working Practices | DHCC +Key Recommendations # The easiest way to reduce your devices’ impact on the enviroment is to use them longer, so use a resource such as iFixit to investigate their longevity and repairability.">Working Practices | DHCC
    Working Practices