Invited Pete Bernard Jason Shepherd Rosina Haberl Michael Kuptz Eric Smiley Woolley, Rob

Summary

White Paper Consensus and Content
The group interpreted the lack of feedback on the white paper approach as consensus and agreed to proceed with the taxonomy-based format. Initial content additions focused on filling out the dimensions table and confirming the standardized usage of the term ‘Edge AI’ throughout the document.

Content Specifics and Strategy Balance
Specific sections were drafted, including engineering services, and the team discussed balancing full-service versus specialized provider strategies. Concerns were raised regarding the document’s heavy focus on telco and networking, prompting a need to better balance constrained, data center, and distributed edge representations.

Finalizing Scope and Normalization
The group decided to omit the Research and Academia section entirely from the current release (V1) to avoid delays associated with academic contributions. The next phase involves finalizing all content submissions and using a scribe or AI tool to ensure cohesive writing styles and formatting

ALIGNED

• White paper taxonomy approach The white paper content approach will utilize the established taxonomy.
• Include more document content More intellectually possible content should be included in the document initially, allowing for refinement and reduction in subsequent weeks.
• Edge AI terminology standard The acronym “E A E” must be replaced with the terminology “edge AI” (all caps) consistently throughout the document.
• Defer Research and Academia content The Research and Academia section on page 18 will be removed from the current white paper version (V1) and saved for potential inclusion in V2 due to contributor availability challenges.
• Normalize white paper writing styles Internal impartial observers, like Linda, or automated tools (AI/Cloud) should be utilized in the final stages to normalize writing styles and ensure the cohesion and readability of the content.

More details:

• Discussion on White Paper Approach and Feedback: The participants discussed the white paper’s approach, which utilizes a taxonomy, and agreed that the silence on the workgroup chat implies violent agreement with the suggested format. Jason Shepherd confirmed they were waiting for feedback before continuing work on the document, and Pete Bernard advised that they should proceed as if there is consensus unless explicit feedback is provided. Rob Woolley noted they had added content, specifically filling out the dimensions table on pages 9 and 10, suggesting that they think more content is better at this stage for later refinement.
• Updates on Document Content and Terminology: Rob Woolley confirmed they performed a search and replace to change the term “AEA” to “Edge AI” in all caps, as previously suggested by Pete Bernard, who expressed appreciation for the change. Eric Smiley reported starting a section on engineering services and consulting, located on page 17, and plans to continue working on and formatting the stream-of-consciousness paragraph they had written. Eric Smiley also proposed including a pictorial representation or chart, possibly in the matrix section, to show the components of different product classes.
• Exploring Service Provider Strategies: The conversation touched on the narrative thread regarding whether companies should choose a provider that offers everything or one that specializes in a single area, agreeing that both approaches have merits. They noted that the decision depends on factors like budget, timing, complexity, and availability, as not many companies can handle everything. Rob Woolley raised a concern that the document is currently heavy on telco and networking, and they sought input on how to better balance the representation of the constrained edge, data center, and distributed edge.
• Revisions to the Dimensions Table: Rob Woolley highlighted two areas in the dimensions table needing review: the power and thermal envelope line, which ties into ML model performance, and the model class section, which attempts to map ML types to locations in the edge continuum. Rob Woolley noted that they updated the power bands with new numbers but needs to find a reference for them and welcomed feedback on the model class attempt. Jason Shepherd committed to reviewing the additions and providing feedback over the next few days, aiming for a significant content push during the upcoming week.
• Contribution for Research and Academia Section: Michael Kuptz inquired about a contributor for the Research and Academia section on page 18, and Rob Woolley suggested VJ, who has an academic background. Pete Bernard cautioned against making this section a dependency for the current release due to academics being busy, suggesting that they should save it for a future version (V2) or engage the academic working group for later contribution. The group agreed to omit the section entirely from the current version.
• Wiki Login Issues: Pete Bernard addressed issues with login information for the Pixelmocker wiki, confirming that some users received it and offered to reissue credentials if needed. Jason Shepherd and Eric Smiley reported receiving emails, with Jason Shepherd noting that their password reset email was entirely in German. Pete Bernard attempted to correct the language setting for Jason Shepherd’s user profile from “site default” to “English US” to resolve the issue.
• Next Steps for Document Finalization: Jason Shepherd plans to provide content and feedback early next week, with the goal of recruiting additional contributors by Friday to finalize the content. Eric Smiley pointed out that once all content is submitted, a lot of co-working or work by a scribe will be necessary to make the writing styles and formatting cohesive. They agreed that Linda, an impartial observer, or an AI tool could be utilized to give the document a read for readability and help normalize the content.

Suggested next steps

• Eric Smiley will continue to work on the section on engineering services and consulting over the next several days to clean up the content, add details about how companies can best utilize services, and consider including a chart in the document to pictorially represent the components that go into different classes of products.
• Jason Shepherd will get some feedback content in early next week and trade notes online.
• Pete Bernard will help reissue login information for the pixel mocker wiki if members need it, and will reset Jason Shepherd’s password and set the default language to English US.
• The group will review the dimensions table offline and provide feedback on the rough shot taken on the model class which talks about the types of ML that might be used in each of the locations in the edge continuum.