Geometrical Thinking Offers a Window into Computation

simonsfoundation.org

72 points·andyxor··9 comments

I'm not seeing an article here, just an offer to sign up to an email newsletter.

Interesting that Mark Churchland is the son of the 'neurophilosophers' Patricia and Paul Churchland. I'm not sure about Patricia but Paul was (maybe still is) a strong advocate of eliminative materialism who also discussed geometric representations of mental representation as points in a multidimensional state space over 20 years ago.

If I went the academic path as opposed to industry I was strongly considering focusing on this:

https://en.m.wikipedia.org/wiki/Information_geometry

> "A clustered representation ... Sherry Choi/adapted from Bernardi et al. 2020"

Why do the authors (of any scientific papers) use a 3-D representation for their diagram? Why are the axis not labeled? It is unclear to me the same diagram could not be constructed in 2-D. Kind'a fluffy.

I think it is not the original figure from the paper but a sketch adapted for general audience? The publishers are now starting to ask scientists to engage audience by "simplified" plots/sketches and keep detailed accurate plots for close community.

> "The publishers are now...'simplified' plots/sketches.."

In fact simplified plots are not new, but rather a common ignorance. On the flip side the "close community" frequently demonstrates their own conceit in their diagrams by frequently failing to go just a small distance more to label their axis. /RANT

Interesting work. I was wondering if you can tell the state of brain activities by looking at these geometric patterns?

seems to me - what's described here probably is this :

here is evidence that localisation applies well even at the level of small neural groups

Reply on news.ycombinator.com