Understanding comes from thinking… actively engaging with ideas. AI may be able to replicate it, but it cannot replace it if it’s a human who seeks to understand.
@garyackerman Surely there is some aspect of brain dynamics and architecture that lead to "thinking" and "understanding".
What makes you so sure that we will not understand those aspects, and then engineer systems based on the same principles?
To quote Daniel Dennett elucidating Stanislaw Lem: "What is the difference between a *simulated* song, and a *real* song?"
@dragonsidedd yeah, that does make me sound like a dualist, doesn’t it!
I’m tooting with a qualitative researcher’s mindset. If I don’t interact with the data myself, then its richness is lost. If we rely on AI to find all of the meanings/ connections and we never look ourselves, then what are we doing at researchers?
@garyackerman What do you do when a typical dataset is giga-, tera- or more-bytes in size, too much for a human even to usefully know what regressions to run, what to focus on and which to discard?
None of our information transmission is flawless. Shannon's entropy applies. High dimensional gradient descent turns out to be the most amazing tool at compressing not raw data, but *concepts*.
Get used to chatting with data that tells you how to understand it.
@dragonsidedd Yeah, we didn’t have that problem years ago :)
We are lucky to have an ever expanding data universe and set of tools.
