So this was a free-rant at the time; I didn’t refine the idea, but I was hoping someone might appreciate the “seed” of thought.
I should clarify when I said, “It’s actually not too complex to understand,” I wasn’t attempting to boast; rather, I meant to convey that it’s a concept more natural than you might realize.
The entire idea of thought as a tree, and not a “tree of thought,” is because “tree of thought” is another concept in philosophy.
My idea was to imagine the branches in a mirror, and yes, even that gets abstract, so I’ll stop there. AI is trained on humanity and human behavior. Thus, establishing a pattern is the difference between emulating one aspect or another.
Because the patterns of generation are based on what is presented, and its training includes learning what to say and when to say it.
So essentially, you’re not only contemplating the sum of humanity in its training, but also considering who says what and when, and how the outliers communicate when it comes to humanity.
Was this any better? If not, I’ll come back next month and see how I’ve refined it.
I’ve overwritten the previous thought because it was completely nonsensical.
Can you clarify with an example?
Starting with a problem exploration both user and model need to understand and agree on what the actual problem is.
Then you have a problem definition.
Based on the definition you start exploring the possible solutions. Select one and you have a solution suggestion. Then iterate over the possible ways to implement a solution.
Select the appropriate approach.
Summarize, write a project document, whatever fits your task at hand.
Feed the summary back into the model asking for analysis and evaluation. Improve the document on the feedback.
Now task the model with implementing a solution to the problem in the defined and specified way.
That’s what I understand from OPs post. It pretty much resembles standard process for project work but with a LLM instead of a whole team.
Hope this helps.
I imagine a thought as a point cloud that changes over time - like a sequence of multiple connections of your brain cells (although inside the brain cells there is alot going on as well, I suppose) and they are played like a small video whenever needed (biochemically activated by hormones).
You might want to have a look into graph databases and LIDAR systems.
At the time no. But i refined it a little more in a new post.
Its sorta like if you think about what is not thought. Then you have yourself what a computer does, then when you think about what it cant do explicitly… when you can answer that, then you have your first step.