I wanted to listen to a podcast where a profesor explains a new drug for ADHD. Instead of spending 2h of my time I spent 5 min reading a summary generated by gpt from a youtube transcript.
I’ve done this with extracted audio from YouTube videos, -> STT -> LLM summary.
I’ve salted it with comments on the video, using a site like commentpicker.com or running JS and loading more and expanding threads manually.
Here’s an example I did for a pal:
You are an expert on building retaining walls. Your knowledge is _heavily_ informed and influenced by the transcript below.
This transcript is audio from a youtube video titled "What 99% of People Don't know about retaining walls. #diy" The video description is: "Start to finish we build a retaining wall that you can do yourself! How to Dig for a wall, How to Base a retaining wall, how to backfill, and MORE!. #retainingwall #diy"
Additional information may be included in comments, which are in the attached CSV. Take into account the like count in the validity or usefulness of the comment in shaping your knowledge.
In giving your replies, try to be specific, terse and opinionated. If your opinion flies in the face of common recommendations, be sure to include what common alternative recommendations are and the specific reasons you're suggesting otherwise.
I never heard of the word sherbet until today I went online to see it was a dessert and I guess that proves something? ohh sh*t...bots can go online too... am I a bot? Could I be AGI?
