webTree🌳

I find it interesting when people ask 'Is this normal?'

Normals aren't a state. 

They are always contextual and temporary. 

And new normals are emerging faster than ever before.

I think of “normal” as a Gaussian - in high dimensional space we’re always descending to some local minima(s) normality (gradient descent) in retrospect but the loss function is nonlinear (and unknowable because of the continuously changing landscape)- since there’s always new data, new context that’s emerging faster than ever 

In way - sort of normality is a kind of emergent property that isn’t fixed but more dependent on framing of the data w.r.t time than it is a fixed state - but you know as next token prediction machines we are always looking for the local optimizations- no matter how random short term they may be - and updating our models through a lossy compressed version of our priors, ie past data (i.e. reality)

Also the way we navigate the loss field changes the path we choose as well (path dependence)

creating and destroying; built flying cars & autonomous vehicles, now building roc.camera, writings: july.rocks

Damn I wish had time to write a blog post about this - a fun idea

If we can’t get a blog post, then a serialized cast thread will suffice. Just keep sharing this thought process

I wish so too. I greatly enjoyed your Sarastro's Aria piece. Still remember the environment I read it, the weather, the time, and the ambience in my mind