2024.03.14

[seanmcilroy29]

---

2024.03.14 Agenda/Minutes

---

Time 1600 (GMT) - See the time in your timezone

• Co-Chair - Henry Richards (WattTime)
• Co-Chair - Naveen Balani - (Accenture)
• Co-Chair – Abhishek Gupta (BCG)
• Convener – Sean Mcilroy (Linux Foundation)

Antitrust Policy

Joint Development Foundation meetings may involve participation by industry competitors, and the Joint Development Foundation intends to conduct all of its activities in accordance with applicable antitrust and competition laws. It is, therefore, extremely important that attendees adhere to meeting agendas and be aware of and not participate in any activities that are prohibited under applicable US state, federal or foreign antitrust and competition laws.

If you have questions about these matters, please contact your company counsel or counsel to the Joint Development Foundation, DLA Piper.

Recordings

WG agreed to record all Meetings. This meeting recording will be available until the next scheduled meeting.

Roll Call

Please add 'Attended' to this issue during the meeting to denote attendance.

Any untracked attendees will be added by the GSF team below:

• Full Name, Affiliation, (optional) GitHub username

Agenda

• Approve agenda
• Approve previous Meeting Minutes

Use case Submission Review

• CAST - SCI to measure the improvement in CO2 emissions when decarbonizing software

GreenAI - Discussion Continued - Henry / Thomas

• Green AI project briefing, which includes:
• Next Steps
  Green AI Project Proposal Early Brief.pdf

WG Project - Dashboard

Project Review updates

• Transforming Organisational Software Sustainability - Pindy

  • Project to be aligned with;
    • Maturity Matrix with the project.
    • Rubric project donated by Indeed

• Software Carbon Efficiency Rating - Chris

  • Baseline Spec
  • Spec Development Proposal and Focus Areas

SCER Working Group (WG) is exploring the use case for SCER standard for LLM (Large Language Models) in generative AI.

• Real-Time Cloud - Adrian
  • Defining the GSF Real-Time Cloud Renewable Energy Percentage spreadsheet.
  • collaboration documentt to create the initial header explanations.

ISO - Sean

• ISO placeholder now available, before its publication- https://www.iso.org/standard/86612.html

Articles

For Review

• Lifecycle Assessment (LCA) for Software Applications using SCI and Impact Framework - Srini

Recently Published

• GSF Endorses AI Environmental Impacts Act

Note:
WG use case template submission - After submitting this issue, your use case will be submitted to the WG Agenda for discussion.
Article submission—Once you submit this issue, it will be assigned to the GSF Editor for review.

Future meeting Agenda submissions

• If you have any Agenda submissions, please let Sean know via email at [email protected] or open an Issue

Next Meeting

• 28th Mar

The US is now on daylight saving, and Europe will move to daylight saving on 31 March.

Adjourn

• Motion to adjourn

---

Meeting Action Items / Standing Agenda / Future Agenda submissions

• Members - Review the Green AI principles document before next Thursday's meeting
• Create a shortened brief version of the green AI principles document and include it in the following newsletter
• Ping Urs to see if he has completed the article on billing principles
• Thomas - Add the Green AI core team meeting to the calendar and distribute the Zoom link

Standing Agenda / Future Agenda submissions

• Standing Agenda Item: The First meeting of each month is a member Introduction / Buddy-buddy meeting to help new members understand the work in the SWG.
• Standing Agenda Item - First meeting of each month - 5 - 10 min round up of what is happening across the other WG

[navveenb]

Attended

[GadhuNTTDATA]

Attended

[marcoshidalgonunes-avanade]

Attended

[srini1978]

Attended

[PindyBhullar]

Attended

[Henry-WattTime]

Attended

[MichaelMULLER]

Attended

[pgregrivera]

Attended

[tmcclell]

Attended

[seanmcilroy29]

attended

[seanmcilroy29]

MoM
Henry opens the meeting at 1600 (GMT)

Green software development and carbon emissions reduction.
Greg Rivera and Michael Emmanuelle discuss their product Cast Highlight, which analyzes source code to identify inefficient and carbon-emitting software patterns. Researchers used the SCI formula to calculate carbon emissions savings after optimizing the application code. Member highlights three specific patterns for improving application performance and reducing emissions, with ten occurrences.

Improving AI model efficiency and carbon awareness.
Greg confirmed that hardware efficiency and carbon awareness were considered in the final calculation for ACI experiments.

Software carbon footprint measurement and reduction.
The experimenters are hesitant to make any conclusions regarding Azure or Amazon because of a lack of alignment with their goals. Pindy suggests that providing information on both hardware and software used for SCI calculation would help in increasing transparency and understanding of the baseline and improvement. Pindy also suggests using the green label application comparison in the SCR project group, which could potentially lead to a broader discussion.

Efficient code patterns and their impact on energy consumption.
Tammy finds the inefficient code patterns interesting and wants to know more about their impact on hardware.

Improving application performance and reducing carbon emissions.
The group discussed how to interpret data on improved function duration after fixing ten deficiencies and predicted a potential 5% improvement with all fixes applied. The study found a 5% improvement in execution duration after fixing Green deficiencies, leading to a potential reduction in CO2 emissions.

Carbon emissions estimation using software.
Tammy asks about the AI used for the carbon EQ plugin project and how it was developed.
Tammy asks about using three years as a baseline, and Greg explains that they used four years as a default value in their calculation. Greg shares that they found a reliable statistic on cloud environments, which they used as a proxy for reduction in energy consumption. The company's software estimates carbon emissions savings by analyzing source code and augmenting with other factors. The team tested their application on different hardware configurations to measure performance improvements.

Sustainable energy and carbon neutrality in software development.
A company in Germany is exploring an alternative approach to software development to address concerns about CPU utilization. Greg suggests using carbon-aware SDK for energy collection, but it's not yet implemented.

Renewable energy transition and carbon footprint.
During a meeting, Greg made sure to avoid oversimplifying the timeline for the renewable energy transition. He clarified any confusion that may have arisen. The marketing team plans to share a case study with GSF using a new process that includes technical and publicly accessible documents. During the same meeting, Henry and another speaker discussed a food platform experiment that uses AI to automate fixes and measure consumption. Henry asked for input on pressing project updates and mentioned an LCA document. However, due to time constraints, the conversation had to be cut short.

Lifecycle assessment and software applications.
The document compares LCA and NCAA methods for software emissions assessment, with a focus on sustainability in AI development. Henry expresses gratitude to everyone who read and commented on an article regarding the measurement of the energy and emissions intensity of AI books. The article received some exciting concepts and endorsement from the Senate. Furthermore, Data Architect Phil Hudson discusses the creation of consumption and utilization calculators for UBS on-prem systems. This process aims to identify hotspots of end-consumption and optimize service usage.

Action Items
[ ] Share detailed calculation slides from Cast case study with the working group.
[ ] Send calculation slides to Sean for case study publication.
[ ] Reach out to Namrata to schedule an interview and overview of the Cast case study for publication.
[ ] Share the LCA mapping to software applications draft article with the WG for review before publication.