12 comments

[ 3.1 ms ] story [ 22.9 ms ] thread
Which NN architectures could be sufficient to simulate the entire human brain with spreading activation in 11 dimensions?

- citing the same paper: https://news.ycombinator.com/item?id=18218504

NetworkX does clique identification [1] in memory, and it looks like CuGraph does not yet have a parallel implementation [2]

[1] https://networkx.org/documentation/stable/reference/algorith...

[2] CuGraph docs > List of Supported and Planned Algorithms: https://docs.rapids.ai/api/cugraph/stable/graph_support/algo...

> They found that groups of neurons connect into 'cliques', and that the number of neurons in a clique would lead to its size as a high-dimensional geometric object.

So cliques of 11 neurons have been found in the brain? I'd hardly call this 11 dimensional.

Does that mean this "4D chess" I keep hearing about is just playing chess with a clique of four neurons?
I am guessing that they can't actually mean 11 spacial dimensions. Are they saying something analogous to multi-dimensional data structures? Like these things exist in 3D space but they are akin to arrays of arrays?
> multi-dimensional data structures

The lowly byte is 8 dimensional in the Boolean domain.

I wonder if it's more like 11 "dimensional analysis" dimensions, rather than space/time+ dimensions. So thinking about something complicated could mean we can process/hold 11 separate variables at a time?
Maybe they're counting "herbs and spices" as dimensions.
String theory confirmed? ;)
jury still out on string theory and supergravity.

that being said…

julian casablancas confirmed.