The twinkle should become a bright LED flashlight at the very least.

If you dig out the paper I submitted to AGI-2016,
you see a extremely terse outline of reasoning leading to
a geometrical object as a limit form of assumptions
described in earlier AGI-20xx and BICA-20xx papers.

That manifold has been well studied by the mathematical community.
Its how you think the hypergraphs are organized that I believe
is the key.

A couple of years ago, I mentioned the use of algebraic topology
in this AGI context to my old dissertation advisor, who promptly
went to NYU's Bobst library to check out the techniques I was suggesting.
At our next meeting, he was saying "that stuff is unintelligible".

Meanwhile, WS would like to see a chapter of the book.

Cheers, Gene


On 3/16/2017 6:11 AM, Ben Goertzel wrote:
Just for general bemusement ... here are some recent rough outlines I
wrote indicating what I think are some interesting research directions
related to hypergraphs, category theory and AI ...

https://arxiv.org/abs/1703.04361

https://arxiv.org/abs/1703.04382

https://arxiv.org/abs/1703.04368

I am also musing on this paper

http://geometry.caltech.edu/pubs/DKT05.pdf

and wondering if we can somehow use it to do some interesting
differential geometry on hypergraphs.   A weighted hypergraph is sort
of like a field on a simplicial complex, right?   A weighted, typed
hypergraph is sort of like a vector field on a simplicial complex.
So maybe differential geometry and algebraic topology etc. on
simplicial complexes can tell us something about weighted labeled
hypergraphs, which in OpenCog are used as the core cognitive
representation.   But this is just a twinkle in my eye at the
moment...

-- Ben




-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to