0 AI concepts - terrytaylorbonn/auxdrone GitHub Wiki

25.0214 (Gdrive)

I use (unintelligent) AI where its a useful and practical tool (just like using search and code-complete) that greatly assists real (human) intelligence.

• Phase #1 Dev (ecosystem) (Copilot). Use Copilot (thats obvious).
• Phase #2 Doc org/content. Use LLM "reasoning" (on the right in diagram below) to research/collect/create initial doc content. This where the tech writer, product owner, marketing etc write the text/diagrams.
• Phase #3 Chat (MAQ/LLM/RAG). Use LLM "chat" (on the left in diagram below) (I've been working on learning-by-demo).
  • Use LLM/RAG and the "scraped" text from phase #2.
  • Create a "MAQ" (my asked questions) with LLM+RAG to create a Copilot-style chat bot.

That last point shows the immense value of AI that I have discovered while putting together the AI API sandbox. AI helps extract the useful content and give it to me in "MAQ" (my asked question) style: My questions answered specifically (including recommendations for my code). I don't have to spend hours/days sifting through endless verbiage that does not show me how to use the product. I simple ask Copilot. Revolutionary. NOTE: AI was not so useful when creating AI drones, because my real challenges were with the open-source-SW / Chinese-component drones that had limited useful documentation for the AI to "scrape".