trained atom space?

[edp...@gmail.com]

Hello community. COG has lots of structures and tools. Are there available pre-trained Atom Spaces and associated training sims?

I am interest in Jeff Hawkins top down predictions versus bottom up perception and mismatches as a source of attention. To start will require something with some level of world model some level of expectations. Being lazy it would be helpful if such a thing already existed. On the bottom up side a sim/environment and perception system would be a first step. Any guidance is appreciated. 

As an aside Googles LaMDA is almost coming to the Atom Space idea. They need to get off the web and have the agent learn its own local book of facts and methods.

Cheers,

Ed

[Linas Vepstas]

Hi Ed,

I don't understand what you mean by "pre-trained".  The agi-bio guys have dumped assorted genetic and protein databases into the atomspace, thus enabling search and crawling and graphical exploration. These are "world models" where the world is biochemical interactions.

I have dozens of datasets of natural languages sliced and diced every-which way.  They are "world models" where the "world" consists of analyzed, digested English text.  I'm trying to ratchet my way up in the sophistication of these models.

The model that I use for natural language, I think it can be applied to sound and vision, or "perception" in general. It is described in the attached paper.  As far as I know, no one has created atomspaces that contain data extracted from image, video or sound sources.

--linas

--
You received this message because you are subscribed to the Google Groups "opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email to opencog+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/3b38a450-a339-4ed0-9cdd-6676f54b0de7n%40googlegroups.com.

--

Patrick: Are they laughing at us?

Sponge Bob: No, Patrick, they are laughing next to us.