After Ethan Mollick’s keynote at SCUP 2024, I decided that I need to spend much more quality time with generative AI tools so that I can better understand how they work and what we can (and can’t) do with them. I still firmly believe that running LLM tools locally is important - if I’m feeding an LLM my content (some of which can’t be shared with third parties just due the nature of my work), I need to know it’s not being shared. The only way to ensure that is by running everything locally. Which means lots of trade-offs.
All of my work basically lives in Obsidian. Notes for meetings. Planning documents of various kinds. It’s all in there. Some of the output gets turned into actual documents, but the work is in markdown files in Obsidian.
I’d tried the Obsidian plugin Smart Second Brain previously, and deleted it because I didn’t see how chatting with an AI about my notes could be useful. I figured I’d give it another chance, just to see.
I installed it, fired up Ollama, and installed the latest-greatest llama3.1 LLM. I had Smart Second Brain crunch through my 4,646 notes. It took about 8 hours. And then I could ask it about my notes - exciting!
And then I got API errors. I figured maybe it didn’t work well with llama3.1, so I deleted the data and reverted to the default nomic-embed-text and llama2 models and reindexed. That only took about 30 minutes. Progress! Here’s the result:
🤖 - AI-Generated Content via llama2
Assistant Hello, I am your assistant. How can I help you?
User What are the most important items from my SCUP 2024 Notes?
Assistant Hello! As your assistant, I’m happy to help you with your query. Based on your existing knowledge enclosed in the XML tags, here are some of the most important items from your SCUP 2024 Notes:
- [[Creative Commons Licenses|creativecommonslicenses]]: These licenses allow creators to give permission for others to use their work under certain conditions. (Note: This is a key concept in understanding how open-source knowledge works.)
- [[Open Educational Resources|openeducationalresources]]: These are educational materials that are made available under an open license, allowing anyone to access and use them. (Note: This is relevant to your interest in SCUP 2024 Notes.)
- [[OER Communities|oercommunities]]: These are groups of people who create, share, and use open educational resources. (Note: Understanding the communities involved in OER can help you better understand the notes you have.)
- [[Open Access|openaccess]]: This refers to the idea that research and publications should be freely available to everyone, without any restrictions such as payment or subscription. (Note: This is an important aspect of open-source knowledge.)
- [[Open Innovation|openinnovation]]: This refers to the idea that innovation can happen through collaboration and sharing, rather than through proprietary ownership. (Note: This is a key concept in understanding the nature of SCUP 2024 Notes.)
- [[Licensing|licensing]]: This refers to the legal agreements that govern how creators give permission for others to use their work. (Note: Understanding licensing is important for working with open-source knowledge.)
- [[Attribution|attribution]]: This refers to the practice of giving credit to the creator of a work when using or sharing it. (Note: Attribution is an important aspect of open-source knowledge, as it helps to ensure that creators are recognized and credited for their work.)
- [[Sharing|sharing]]: This refers to the act of making resources available to others, either through a formal process such as publication or through informal means such as sharing with colleagues. (Note: Sharing is an important aspect of open-source knowledge, as it allows for collaboration and the creation of new works.)
- [[Collaboration|collaboration]]: This refers to the act of working together with others to create something new. (Note: Collaboration is a key aspect of open-source knowledge, as it allows for the creation of new works through the sharing of ideas and resources.)
- [[Community|community]]: This refers to a group of people who share common interests or goals. (Note: Understanding the community involved in SCUP 2024 Notes can help you better understand the notes themselves.)
I hope this helps! Let me know if you have any further questions.
Which has absolutely no connection to what actually happened at SCUP 2024. I have no idea where it got those “most important items” from. It’s linking to notes that don’t exist, for topics that weren’t discussed at SCUP.
For comparison, I fed my “SCUP 2024 Notes” note into ChatGPT-4o, and got:
🤖 - AI-Generated Content via ChatGPT-4o
Here are the most important points from your SCUP 2024 conference notes:
Keynote: From Disruption to Opportunity: Embracing the AI Revolution
- Presenter: Ethan Mollick
- Overview of Generative AI: Highlighted the current state of generative AI in higher education and its transformative potential.
- Live Demos: Demonstrated real-time applications of AI tools like Claude, Google Gemini, ChatGPT-4o, Midjourney, and AIVA.
- Impact on Education: Showed how AI can drastically reduce project completion times and emphasized the need to integrate AI tools into teaching and learning processes.
How to Design Neuro-Inclusive Environments
- Presenters: Scott Montemerlo, Allison Page, Michael Tyre
- Learning Spaces: Comparison between conventional and active learning classrooms, highlighting the importance of flexible, inclusive learning spaces.
- Design Principles:
- Mitigate distractions (noise, glare, patterns)
- Provide choice and control over the environment (lighting, privacy, furniture)
- Balance flexibility with predictability to reduce anxiety.
- Sensory Overlays: Use sensory mapping to identify zones within learning spaces.
Integrating the Planning of Learning Spaces & Teaching and Learning Strategy
- Presenters: Tracey Birdwell, Chris Morett
- Alignment: Emphasize aligning learning spaces with the institutional teaching and learning strategy.
- Ongoing Conversations: Importance of continuous engagement and benchmarking in planning.
- Community Development: Engage faculty through programs like the MOSAIC Fellows program to understand and advocate for their space needs.
Using Occupant Evaluation Data to Inform Higher Education Design
- Presenters: Joe Klocek, Jeff Blaho
- Post-Occupancy Evaluation: Conduct evaluations six months after move-in to understand usage and satisfaction and document changes from previous environments.
Lessons Learned from a Successful Strategic Planning Process
- Institution: Washington State University
- Challenges: Leadership transitions during planning and implementation.
- Planning Software: Use software like MS Viva Goals, achieveit, cascade, and others to track progress and generate reports.
- Communication: Regular updates and multiple communication channels to keep the community informed.
Engaging Faculty in Classroom Planning & Design
- Presenters: Karen Detlefsen, Bonnie Netel, Gary Purpura
- Central Pool of Classrooms: Involvement of faculty in planning and decision-making for classroom design and usage.
- Subcommittees: Four subcommittees focus on utilization and scheduling, technology, operation and maintenance, and design and renovation.
- Education Methods: Classroom tours, documenting pedagogical approaches, and sketching layout designs for discussion.
Planning for Neurodiverse Campus Communities
- Neurodiversity Awareness: Understand and design for neurodiversity, considering sensitivity to stimuli, need for movement, and privacy regulation.
- Design Considerations: Incorporate features like varied sensory inputs, flexible spaces, and wayfinding aids to support neurodiverse students.
Relationship Building: Strategies for Inclusive, Student-Centered Design
- Presenters: Christopher Gluesing, Kathleen Kugi-Tom, Katherine Nichols, James Pfeiffer
- Inclusive Design: Importance of field trips to other institutions, common knowledge building, and collaborative design processes.
- Resources: Books and guidebooks on inclusive design strategies, such as “ALL: A Guidebook of Strategies for Inclusive Design.”
Exploring History of Place Through Integrated Planning to Create an Inclusive Campus
- Presenters: Yvonne Choe, John Souleles
- Case Study: UCalgary Faculty of Veterinary Medicine Spy Hill campus, focusing on integrated planning and inclusive campus development.
- Hiring Curriculum Specialists: UCVM is hiring curriculum specialists to support program and course adaptation.
Which is an actually useful summary of the highlights from my notes.
OK. Maybe it’s something funky with Smart Second Brain and/or its connection to Ollama. So I tried a different LLM plugin for Obsidian: Smart Connections. It has a built-in LLM, BGE-micro-v2, which is much smaller but is self-contained. Indexing took about 10 minutes. But it uses a different LLM to do the chat interface, and doesn’t come with one built-in? So I set it up to use Ollama and phi3.
It looked promising, so I tried:
🤖 - AI-Generated Content via phi3
system
Work/practice/conferences/SCUP/SCUP 2024 Notes.md
user
summarize the important items in [[SCUP 2024 Notes]]
assistant
API Error. See console logs for details.
Likely something funky between the plugin and Ollama. I’ll try to debug that later. Console logs? Where the heck are they? Can’t seem to find a relevant Obsidian log file anywhere…
BUT - there is a third plugin - Smart Connections: AI-Powered Note Connections - that crunches notes as you open them and uses the LLM data of my entire notes vault to find related notes. This could be useful! And, it actually works, using the built-in BGE-micro-v2 LLM. But, is it something that would be useful? It opens a sidebar that lists possibly-connected notes, ranked by relevance. Here’s what it looks like:
Screenshot of the Obsidian Smart Connections sidebar, listing related notes
This is the kind of thing that might be useful - not chatting with some LLM about my notes, but having something digging into my Obsidian Vault to find things for me. But - this sidebar doesn’t look super useful. Sure, the listed notes might have content that’s related to the note I’m viewing, but so what? I’ll try leaving it running for awhile to see if the suggested connections are actually useful. This might be one of those things where it sounds exactly like what I’m looking for (an automated assistant mining my notes to help me organize and make sense of them all? amazing!) but just doesn’t hit the mark (is listing notes really helping?)
The Smart Connections plugin also has a graph view based on the LLM crunching of the vault. Which sounds like it would be awesome, but I just got an error message saying there were no nodes to load or something. Helpful.
Anyway. Spending some quality time with LLMs, trying to figure out how/if/why they might be actually useful to support my work. I’ll add a follow-up Note if I ever figure out how to get the connection between Obsidian and Ollama working properly.