Google's NotebookLM analysis of this blog

I am having too much fun! I put all the blog posts from this blog into Google's NotebookLM [ https://notebooklm.google.com/ ]. Here is what NotbookLM gave as a  summary:

"James C. March explores a wide range of philosophical and practical topics, arguing that non-violence involves creating options for others and that creativity is a form of responsible non-violence. The author questions the prevailing economic focus on efficiency and profit, particularly in industries like pharmaceuticals, suggesting it often disregards human well-being and democratic values. A central theme is the evolution and propagation of "memes" (ideas and culture) versus "genes" (biological survival), positing that human focus is increasingly on memetic survival through shared knowledge and collective understanding, facilitated by communication technologies like the internet and practices such as the scientific method for evaluating perspectives. The text also reflects on personal experiences, the nature of emotions, the challenge of learning, and the search for a universal philosophy of life grounded in the fundamental principle of survival." 

Not too bad for a summary of 20+ years of blog entries.

I used NotebookLM to create an "Audio Overview", which I published on Spotify [ https://open.spotify.com/episode/43mE3cJ5iYyRDFSBB79yaL ].

I've been trying slightly different versions of the input document, and getting some unexpected results in the corresponding "Audio Overview":

1. The order of the blog entries makes a difference.

I first submitted a document with the blog entries in reverse date entry, most recent blog post first. The audio overview focused too heavily on the recent posts. I reversed the order, putting the posts in date order. This seemed to give a more equal weighting to the posts. Note: The audio overview I published on Spotify is based on the date order posts.

2. How I referenced the blog entries matters.

A few versions ago, I submitted my blog posts without proper reference to each blog post. I didn't label each as written by me, or that all the posts came from the same blog. The resulting audio overview was strangely boring, and it wasn't just because the audio never mentioned my name (referring to the blog posts as "your sources"). In particular, the audio overview never knew the posts were all from the same person. There wasn't as much effort to try and tie the posts together. In retrospect, this makes sense, but it still surprised me.

3. WARNING! The exported XML from Blogger will include drafts of unpublished blog posts. I had to edit these drafts out after they showed up in a podcast, much to my surprise!

Overall, I'm blown away, by the two manufactured voices, and the summarizing of so many diverse blog posts over the past 20 years! Never in my wildest dreams did I think I would see such powerful artificial intelligence in my lifetime!