François CholletVerified account

Reminds me of the book The Denial of Death by Ernest Becker. How he pinned down every human misery onto the single idea that we don't accept death and find ways around it.

Also a story of publishing innovation. I think ‘Cod’ was the first book in the line and it started a whole genre. https://archive.nytimes.com/www.nytimes.com/books/97/09/21/reviews/970921.21benjamt.html …

And, unfortunately, we often assume a linear relationship

ATOMS: HOW THE CAUSE OF ALL MAN'S PROBLEMS CAN TURN INTO THE SOLUTION TO ALL OF THEM OR VICE VERSA BE CAREFUL

It also makes researching (if not actually writing) these books a lot of fun. Perhaps that's a vestigial remainder of our love for the rabbit holes that sheltered early mammals for tens of millions of years. I my new book, Rabbit Hole, I will show...

Same with neural networks: there are many adversarial examples showing things like "this single pixel tells me that this is...". I guess both cases are optimized to come up with simple conclusion given complicated input.

Helpful though as a heuristic for giving the universe handles that a finite human operator can comfortably work with, without the other constraint: that of impossible overwhelm. As long as we don't confuse the maps with the territory, it's useful to build knowable maps.

Some of it might be “Apophenia”: tendency to perceive meaningful patterns with random data. I call it cloudxplanations, after apophenia-induced cloud-shape finding.

The simplification does not happen on the learning part of the brain (the right brain if u may) but when trying to explain serially to someone using words.

Sounds like you’re gonna love @kumupowered mapping... https://kumu.io/manifesto