Technical presentation by Fabien Bénétou, moderated by Dene Grigar.
https://git.benetou.fr/utopiah/spasca-fot-sloan-q1/src/branch/main/backend/converters

AI Summary
The meeting covered various aspects of a virtual reality project, including file conversion, text manipulation, and 3D object interactions. Discussions focused on technical challenges, user experience considerations, and potential applications in academic and collaborative settings. The group also explored the role of AI in their work and debated the transparency and interpretability of large language models.
Next steps
- Fabian to continue development and refinement of the VR authoring tool, focusing on user experience and functionality.
- Dene to begin work on integrating sound into the WebXR experience using the newly acquired funding.
- Team to explore ways to output VR-authored content in formats compatible with non-VR users, such as HTML or simplified 2D representations.
- Fabian to investigate the possibility of implementing multiple selections and clipboards in the VR environment, as suggested by Peter.
- Team to consider implementing visualization and manipulation of links between documents in the VR space.
- Leon to explore the potential use of SVG format (with possible extensions for 3D) as a common denominator for exporting spatial arrangements from VR.
Summary
AI Tools and Sound Integration
In the meeting, Fabien and Dene discussed the possibility of adding weight to their water activities and the need for better tools. They also talked about the upcoming presentation on file conversion and the potential for a lengthy discussion. Dene mentioned a new idea from Frode, but decided to wait for him to present it. Fabien shared his progress on sound integration into the Webex R experience and mentioned the need for strong feedback during sports. Dene mentioned receiving funding for sound integration and offered to share the code with Fabien. They also discussed the various AI tools they are using.
Balancing Technology and Content in Demos
In the meeting, Mark and Fabien discussed the importance of striking a balance between technology and content in their work. Fabien emphasized the need for clear communication during demos and the importance of understanding the context of the technology being demonstrated. They also discussed the challenges of presenting complex technology to non-experts and the need for a flexible architecture that can accommodate different domains and file formats. Dene facilitated the meeting and ensured that Fabien’s screen was visible to everyone.
Open-Source Converter Deployment and Security
Fabien discussed the open-source nature of a converter, emphasizing that users can modify and deploy it as they see fit. He clarified that the converter runs on the back end, handling heavy computations like converting PDFs, which can be time-consuming for large files. Peter raised concerns about running the back end in a shared hosting environment, where processes longer than 1.3 minutes might be terminated. Fabien reassured him that the setup could be deployed on a dedicated server or even at home, and provided instructions for setting it up. Dene emphasized the need for discretion during the development phase.
Project Open Access and Hardware Requirements
Dene discussed the project’s open access and the need to avoid sharing it with big corporations. Peter suggested a workaround for shared hosting environments, and Mark emphasized the importance of understanding the limitations of the headset. Tom asked about hardware requirements for running servers, and Fabien explained the synchronization of files and the need for converters. Dene clarified that the project is currently open access and will likely follow Creative Commons Attribution Share 4.0 International. Fabien discussed the minimalistic nature of the project and the potential need for more resources depending on the type of content.
PDF to XML Conversion Process
Fabien discussed the conversion process of a PDF to an XML file and images. He explained that the conversion is done on the backend and the resulting file can be manipulated in XR. Fabien also mentioned that the conversion process is not lossless, meaning some modifications may not be retained. Dene suggested that a two-way conversion might be useful in the future, but Fabien clarified that the goal is not to create a PDF authoring tool. Peter suggested using bidirectional transformations and partial lenses for this purpose.
Text Collaboration in Virtual Reality Environments
The group discusses the challenges and goals of working with text in virtual reality environments. Mark emphasizes the need to move away from legacy thinking about document editing and focus on structured data exchange. Tom stresses the importance of being able to extract and share the results of collaborative work in VR. Dene clarifies that the current focus is on text annotation and manipulation rather than text creation. The group explores ideas for output formats, including XML and interactive web representations. Peter suggests developing more intuitive ways for academics to work with LaTeX in VR. Mark highlights different modes of collaboration and the importance of sense-making for individual users. The discussion concludes with Dene emphasizing the need for “born virtual” writing tools rather than adapting print-based structures to VR.
VR Text Manipulation and Spatial Organization
Fabien demonstrates a new feature in VR where users can manipulate lines of text or 3D objects by pulling them through rings, applying transformations such as color changes or scaling. The system allows for composable transformations and stacking of multiple effects. Fabien suggests this could be useful for applying complex packages or academic publishing tools. The group discusses the potential of spatial organization for summarizing and presenting ideas, with Leon and Tom expressing interest in using VR to create spatial representations of complex concepts. There is debate about the role of AI in the process, with some preferring manual control while others see potential for AI assistance.
LLMs Transparency and Interpretability Challenges
The group discusses the challenges of making large language models (LLMs) more transparent and interpretable. Karl argues that it’s not feasible to “unbox” LLMs due to their complexity, while Tom suggests it may be possible with smaller, limited models. Dene proposes applying LLMs to smaller units of information, such as a faculty manual, to create curated knowledge systems. The conversation then shifts to Fabien’s work on file manipulation and visualization in a 3D environment, with suggestions for improvements like multiple selections and linking related documents.