Please note that I've worked in both biology and CS for 30+ years. Everything I describe in this way is a simplified approximation of mental processes, and half-joking (mainly appreciated by ML nerds and cognitive neuroscientists, I admit)
To build an embedding for a human means to build an approximate map of entities within a metric space. We do this all the time, for example unsupervised clustering of related entities (for example species classification based on visible character states).
From what I can tell, the human brain basically has a number of approximate computing systems- we have associative memory (hash tables/dictionary lookups), trajectory interpolation/extrapolation (catching a ball with only a few glances), logical reasoning from a premise (theoretical math), monte carlo tree search (formulating a complex plan wiht multiple steps and some unknowns), etc, etc. I certainly don't propose that my terminology is specifically scientifically correct, it's just heuristics I apply to my own thought patterns.
> To build an embedding for a human means to build an approximate map of entities within a metric space. We do this all the time, for example unsupervised clustering of related entities (for example species classification based on visible character states).
I thought you said "for a human". These are all computational concepts that are unrelated to human cognition (at least in terms of evidence). Why are we talking about human cognition in terms of "metric space" when metric space applied to human cognition is nonsensical?
> From what I can tell, the human brain basically has a number of approximate computing systems- we have associative memory (hash tables/dictionary lookups), trajectory interpolation/extrapolation (catching a ball with only a few glances), logical reasoning from a premise (theoretical math), monte carlo tree search (formulating a complex plan wiht multiple steps and some unknowns), etc, etc. I certainly don't propose that my terminology is specifically scientifically correct, it's just heuristics I apply to my own thought patterns.
Police? This comment right here: they failed to read my comment.
I of course certainly strongly support you describing your mental cognition the way you want to! It's just not evidence that other people can be modeled with the same computational mechanics, nor that these mechanisms are sufficient. We should use meaningful terms that people can actually recognize rather than trying to forcing mental cognition to fit models that are curently in vogue.
To build an embedding for a human means to build an approximate map of entities within a metric space. We do this all the time, for example unsupervised clustering of related entities (for example species classification based on visible character states).
From what I can tell, the human brain basically has a number of approximate computing systems- we have associative memory (hash tables/dictionary lookups), trajectory interpolation/extrapolation (catching a ball with only a few glances), logical reasoning from a premise (theoretical math), monte carlo tree search (formulating a complex plan wiht multiple steps and some unknowns), etc, etc. I certainly don't propose that my terminology is specifically scientifically correct, it's just heuristics I apply to my own thought patterns.