I Just finished an interesting exercise in AI.
City Council meeting minutes dating back to January 2021 uploaded to 1.5 pro in the Google AI studio (over 100 .pdf documents condensed and converted into 4 text files, Over 200,000 words. Each text file contained 1 year of minutes in plain text format).
Instructions (very condensed): Only report on agenda items that had votes in conflict (no need to show consent items, or unanimous votes). Then format the text into outline format.
This was to provide the residents of my dear city the voting record of their city official seeking reelection.
I used about 800K of the 1 Million tokens including the source files to complete.
End result: https://www.thepalmbayer.com/p/palm-bay-city-council-voting-record.
* It's a fun page to talk to with an AI extension also. Now that I have this large data set that knows each members voting pattern over 3+ years I wonder if it can predict...........
** Gemini 1.5 pro with a 1 Million token context window is a beast.
[link] [comments]