Joscha Bach Universe
Try for yourself on https://joscha-bach-universe.fly.dev/
Why I created the Joscha Bach Universe
The first time I listened to a podcast of Joscha Bach was on 2023-09-04. I remember so precisely because it was so fascinating that I started taking notes, so I could better understand and remember his ideas. His worldview gave clarity in area’s that where foggy for me. It resonated with the philosopher in myself. But he didn’t just stick with the philosophical explainations, he also shared personal experiences about his background, loneliness in childhood, and personal and relationship growth. This resonated with the therapist in myself.
In July 2024, after a psychedelic session, I had a special experience where it seemed my consciousness switched off and rebooted itself, experiencing how I started distinguishing somehting from nothing, then seeing opposites like darkness and light and gradually making more sense of the world I live in by giving meaning to my experiences. I could even visually observe emotions and sounds, realizing my experiences are brain-made. This profound experience reminded me of Joscha’s explanations, and inspired me to listen to Joscha Bach even more.
Listening to more podcasts, I took extensive notes and began transcribing, annotating them in my Zettelkasten in Obsidian. This started to feel like biblical study – a lot of work, and as someone naturally unstructured, it required energy.
I tried asking ChatGPT and other LLMs. While responses were often good, sometimes they were inaccurate, mixing other thinkers’ ideas with Joscha’s. I wanted a precise solution with exact podcast citations to validate responses directly.
This is why I started to build this chatbot. To have a precise and exact source for each response. And this resonated with the developer in me. Although it’s been a long time ago that I did some proper coding, I started to learn python, react, and LLMs. And here we are, with a first version that fulfulls some of my needs. I see many potential improvements, but it already has the ability to give me more insight, with the ability to see the source of each response.
Why is creating important?
The chatbot will give a nice answer to the question “Why is creating important?”
-
Empowerment: Creation empowers individuals by allowing them to shape their own experiences and environments. It fosters a sense of independence and reliability, as one can depend on what they create and have the ability to recreate it when necessary (Source ID: 1, 6).
-
Meaningfulness: Engaging in creative activities often leads to a more meaningful life. Many people find that their lives are most fulfilling when they contribute to the world through creative endeavors, rather than merely performing administrative tasks or consuming experiences (Source ID: 4).
-
Community Building: Creating together with others fosters healthy communities. It encourages interaction and collaboration, which can lead to the development of shared experiences and spaces that enhance social bonds (Source ID: 4, 7).
-
Self-Expression: Creativity is a way to actively shape oneself and express one’s thoughts and feelings. It allows individuals to capture conscious states and explore their inner worlds, which can be intrinsically rewarding (Source ID: 3, 8).
-
Practical Skills: Engaging in creation often involves learning practical skills that can be useful in everyday life. This not only enhances one’s ability to provide for themselves but also makes the process of creation more enjoyable and fulfilling (Source ID: 6).
This response inspired me to create even more projects and start building this website. Check the chatbot for yourself on https://joscha-bach-universe.fly.dev/ and ask it a question.
Technology and Architecture
All technologies I used to build this project were new to me. I had to learn them all from scratch. I was helped a lot by AI, more specifically Cursor and Claude. I learned a lot about python and react.
This is a very basic RAG (Retrieval Augmented Generation) chatbot.
I started by creating a structured list of many relevant podcasts and presentations I found of Joscha Bach. I then used Transkriptor to transcribe the audio of the podcasts and presentations. It separates the different speakers and gives the timestamps per sentence that is transcribed.
Langchain is used to create the RAG chatbot. The transcribed sentences are combined in larger chunks (4 sentences wihtout overlap and 12 sentences with 3 sentences overlap) and stored in the Pinecone vector database to store the embeddings of the podcasts and presentations. Each chunk also contains a URL to the specific podcast or presentation, including a timestamp to the exact sentence.
For each question that is asked to the chatbot, the chatbot first searches the Pinecone vector database for the most relevant chunks. It then uses the OpenAI API to generate a response based on the most relevant chunks.
Since the link to the specific part of the video is stored in the Pinecone vector database, the chatbot can directly link to the source of each response, and you can hear Joscha say the original sentence.