♥️ Join as a Patreon Member to help me make better / more videos - not to mention all the Patreon benefits you get!
[ Ссылка ]
📚 I use Readwise/Reader to capture my daily digital text intake.
Readwise (2mo free trial): [ Ссылка ]
Reader (2mo free trial): [ Ссылка ]
🔗 Useful Links
[ Ссылка ]
[ Ссылка ]
[ Ссылка ]
[ Ссылка ]
Description:
In this quick update, I'm excited to share my new favorite local AI model: Gemma 2 by Google. As someone who's always on the lookout for the best AI tools to integrate with Obsidian and improve our note-taking experience, I was blown away by Gemma 2's capabilities.
I'll walk you through why Gemma 2 has become my go-to model for most tasks, especially when it comes to following directions and accomplishing specific goals. While I still prefer LLAMA3 for creative applications, Gemma 2's concise and accurate responses have made it my daily driver for nearly everything else.
Here are some key points about Gemma 2:
- It's a 9 billion parameter model, but the quantized version is only 5GB
- Runs smoothly on my Mac M1 with 16GB RAM
- Excellent for task completion and following specific instructions
- Performs well in practical, conversational tests
I'll also touch on how I test these models, moving beyond traditional benchmarks to focus on real-world applications, especially within the context of note-taking apps like Obsidian.
For those interested in trying Gemma 2, I'll show you how to download and update it using Ollama. If you're already running LLAMA3 without issues, you should have no problem with Gemma 2.
Stay tuned for more updates as we continue to explore and integrate the best AI models into SystemSculpt and enhance your Obsidian experience!
🔧 Thanks for watching SystemSculpt!
Make sure to like and subscribe so I can keep making this vids for you! Every interaction is appreciated; it helps me out more than you know.
📧 Contact Me!
I read/respond to all my emails! Contact me at systemsculpt@gmail.com.
🧾 Copyright Disclaimer:
The content presented in this video may include copyrighted materials used under the "Fair Use" policy. I do not claim ownership of any third-party content and I acknowledge the rights of the respective copyright owners. If you believe that your copyrighted material has been used inappropriately, please contact me at systemsculpt@gmail.com.
© 2024 SystemSculpt. All Rights Reserved.
Timestamps:
00:00:00 Introduction and Surprise Announcement
00:00:11 Gemma 2 Model Introduction
00:00:26 Approach to Testing AI Models
00:00:48 Understanding LLMs and Pattern Recognition
00:01:11 Practical Testing Methods
00:01:25 Obsidian Knowledge Example
00:01:51 Comparing Gemma 2 and LLAMA3
00:02:07 Model Preferences for Different Tasks
00:02:21 Local Model List and Management
00:02:46 Updating Models with Ollama
00:03:04 Gemma 2 as Daily Driver
00:03:18 Recommending Gemma 2 and Specifications
00:03:26 Technical Details and Hardware Requirements
00:03:44 Conclusion and Hardware Compatibility
#GemmaAI #LocalAIModels #ObsidianAI #AIProductivity #SystemSculpt
![](https://i.ytimg.com/vi/EsCV3-Pn734/maxresdefault.jpg)