I find Logos an indispensable tool, but as a preacher, it's always utterly failed me in one key area: Illustrations.
The tag-based illustration guide is basically just a combined glossary for my books of illustrations. It's nice to have available, but I almost never use it. I have one or two anthologies that I really like, and I only delve into the others if I'm desperate.
The Illustrations generated by Sermon Assistant are much too vague and obvious to be useful, even if I wanted to include AI-generated content in my sermons.
What makes the search for good illustrations especially maddening is that I know my library is full of them. I have sermon collections, church history books, a library of classic literature, and the whole "Sermon Illustrator" set, but it would take years to go through and tag everything that might be useful.
When the Study Assistant was first announced, I was very excited because I thought I would finally have a way of searching for specifically themed illustrations within my library. Unfortunately, that's not something Study Assistant is capable of doing. After experimenting with several different prompts (over several agonizing weeks), I finally got it to admit the following:
"My search function doesn’t allow me to:
1. Filter by narrative form — I cannot systematically distinguish between narrative portions and didactic passages across your entire library.
2. Detect thematic patterns across diverse genres — While I can search for keywords and concepts, I cannot comprehensively scan fictional works, sermon illustrations, and church history narratives simultaneously to identify characters and events matching your specific theme.
3. Locate and extract specific passages — I cannot systematically retrieve the precise textual locations that best exemplify your theme across multiple book collections."
Would it be hypothetically possible to make the Study Assistant capable of performing the above functions?
Would it be possible to update the Sermon Assistant, allowing it to search for illustrations rather than merely generate them?
Alternatively, would it be possible to run a one-time AI-powered process to catalogue content in my library and make things easier to find?
