[ 13/6/20 Tom’s Facebook page]
I understand that you are committed to your interpretation in a way that admits of few alternatives, and I’ll give it another try.
Consider an alternative interpretation of the same phenomena you observe.
Yes – there does appear to be polarity between groups like democrats and republicans, and that polarity can result from an over simplification of a complexity. The simplest model we can make of anything is a binary.
We can take any large collection, even any infinity, and divide it into two. Such a simple division is the simplest possible, and therefore the most likely to occur in any neural network.
Thus people using simple models tend to form polar groupings.
Consider that the space of possible geometries and constructs and relationships is infinitely larger than that, and that is the simplest possible one, and therefore one that can be seen everywhere if one takes the simplest possible interpretation (at whatever meta level one approaches any particular subject).
For me, the reality is very different.
For me, clearly, most things are in reality extremely complex, and what seems simple is so because of the necessity of our perceptual and computational systems simplifying them down to make them computationally accessible within the computational ability and time available.
Goodhart’s Law (from complexity theory) is well worth recursive contemplation in this context of dealing with complex systems – “Any observed statistical regularity will tend to collapse once pressure is put upon it for control purposes”. When you consider this in the context of our legal and cultural and conceptual systems, it becomes clear that all attempts to regulate and codify that we see expressed are a failure to comprehend complexity.
To get any sort of reasonable handle on complexity one must be willing to enter the dance with the system of constantly probing, assessing, responding, and part of the response must be developing new probes and new schema for interpretation of results. Only in this fashion may we successively refine our models and approximations and become successively less wrong.
And it seems clear to me that reality is sufficiently complex that any computational system attempting to model it (human or AI or AGI) will be essentially wrong in important aspects – eternally. Infinities have that unsettling characteristic.
The biggest trap we face is being too certain of our “answers”, because in that certainty we stop looking for alternatives, and if any infinity exists, then logic demands that alternatives exists. And the evidence I have indicates that it is very probable that an infinitude of infinities exist (and that idea can be a might disconcerting when anyone first gets a real handle on it).