Playing with AI
Author: Klaus Mager Issue: 2024-09-04
Playing with AI
by Klaus Mager
I asked Gene Bellinger to host a conversation with myself and Joni Kindwall-Moore, who I have been working with as an advisor and strategist. We are developing what is referred to as an Innovations Brokerage focused on regenerative agriculture, which also includes an IT company specialized in blockchain technology. Gene recorded the conversation:
[Image not included in the current archive. Images may be included in the future.]
Gene has been experimenting with Gemini and other AI models to create Kumu maps documenting recorded conversations. This was yet another experiment, and here the first iteration of our talk:
He then asked Gemini to simplify the labels so the map was less dense and got the following:
https://kumu.io/-/961046#map-vbwg8aSp
I spent some time reading the maps, but couldn't develop a coherent story line that would correspond to the map without dedicating a block of time. As an experiment on my side, I took the map and inserted the first iteration into my chatbot, asking for translation into text. ChatGPT 4o took an annoyingly brief time to create a response developing a spot on story line that I was then able to iterate a couple of steps into a useful summary.
https://chatgpt.com/share/9fbf2337-0962-44ea-b651-637f9edbe6a7
While Gemini is getting quite good creating system maps, ChatGPT 4o is still struggling with that, but has no issue reading and interpreting them. The combination of such specialized AI models to map complex and free flowing conversations into a systems map, then using a LLM to translate into text is compelling, saving time, adding expertise.
The text translation is quite good summarizing key points and recommending next steps. I've been training this chatbot for almost 2 years on food and agriculture related issues, the knowledge base has solidified, cross checked, supported by best available science. I can now imagine using this process to engage with subject matter experts on focused conversations, then using AI to extract key relationships, dependencies, and actionable next steps.
Related:
- Klaus Mager (author)
- 2024 (year)
- Topics: Regenerative Systems and Agriculture