MemoryWalk — AI That Encourages You to Touch Grass

What if artificial intelligence could inspire us to appreciate the world beyond our screens? 🌿 In today's digital age, we often rely on our smartphones to learn, communicate, and pass the time. However, this constant screen time can lead to neglecting the sights, sounds, and sensations that surround us. During the Hacktoberfest Open-Source AI Challenge — Week 1: Touch Grass, I contemplated this phenomenon.

Rather than creating another app that encourages more screen time, I aimed to develop something unique. And so, MemoryWalk was born. <source 97ddaf2e-27b>

**What MemoryWalk Does**

MemoryWalk is a user-friendly web application that harnesses locally-run artificial intelligence. Its primary purpose is to provide real-world observation tasks, prompting users to disconnect from their devices and engage with their environment. The concept is straightforward: AI suggests an observation task, users put their phones away, and then share their observations upon returning. Here's how it functions:

1. Determine the duration and focus of your mission.

2. Receive a mission generated by AI.

3. Put your phone away and go explore your surroundings.

4. Return and answer a few questions about your experience.

5. Receive personalized feedback, a score, and a reflection.

6. Save your experience in a virtual "Memory Garden."

For instance, a mission could task you with becoming a sound detective for ten minutes. Instead of listening to music or mindlessly scrolling, you would concentrate on the sounds of nature—birds chirping, cars passing by, the wind rustling through leaves. Another mission might involve observing the colors and patterns of plants in your vicinity. These tasks are simple, yet they encourage a deeper engagement with the environment.

**The Demo**

To truly grasp the essence of MemoryWalk, it’s best to witness the experience firsthand. The demo illustrates the journey from the initial screen, through generating a mission, completing the reflection process, to saving the memory in the Memory Garden. <source 97ddaf2e-27b>

**How It Works**

MemoryWalk is a local-AI-powered experience that promotes screen-detoxification and mindfulness. Users select their available time and area of focus. The application then generates a tailored observation mission using a locally-run AI model. Once the mission is completed by putting the phone away and returning to reflect, users receive AI-generated feedback and save their experience.

**The Technical Details**

The project employs a simple stack—HTML, CSS, and JavaScript for the user interface, Python for the backend and local web server, Ollama for running the AI model, and qwen2.5-coder:3b for mission, reflection question, and feedback generation. The browser communicates with the Python backend, which forwards prompts to Ollama. The resulting responses are then displayed within the application.

**Why Open Innovation Matters**

Open innovation played a crucial role in this project. Using Ollama, an open-source language model, allowed the creation and testing of MemoryWalk on a personal computer without relying on paid APIs. This approach provided several advantages:

1. **Control over experimentation:** The ability to modify prompts, test various missions, and experiment with different models without altering the entire application structure.

2. **Local-first approach:** By running the model locally, the project minimized reliance on a hosted AI service, thereby enhancing privacy and control over data processing.

3. **Lower barriers for developers:** A local runtime setup enabled individual developers to experiment with language models even without extensive cloud infrastructure.

4. **AI without constant online presence:** This project explored AI's potential to initiate offline activities such as observation, listening, reflecting, and memory retention. It showcased AI's utility beyond content generation, focusing on real-world engagement.

**Insights Gained**

The most significant takeaway was that AI is merely a starting point. The surrounding experience is equally important. A well-structured mission, a natural reflection process, and an interface that encourages screen avoidance are crucial for user engagement. Additionally, attention to technical details—such as seamless integration between frontend and backend—improved the project's usability. Open tools facilitated this exploration and empowered individual developers to build prototypes using accessible resources.

Overall, MemoryWalk demonstrates the potential of AI to encourage real-world experiences while keeping screen time in check. It's a testament to how innovative and practical applications of AI can complement, rather than replace, our offline lives.