On Mon, Jan 27, 2014 at 11:10 PM, Aaron Hosford <[email protected]> wrote:

> How much room does your system leave for alternative ways of doing things?
> Is it a framework with interchangeable modules that conform to a
> standardized interface?
>
>

OpenCog is highly modular; it is a framework such as you describe...

However, modules currently must be written in C++ or python, in order to
operate properly with the Atomspace knowledge store...

If you have your own single AI algorithm and don't want it to interface
with any other algorithms, there isn't really much point to using OpenCog,
unless your algorithm matches very naturally to a "weighted labeled
hypergraph" knowledge representation....  No point to wrap your code in a
general framework unnecessarily...

The value of using OpenCog for AI research comes from

-- the possibility to inter-operate your algorithms with other algorithms
on the same knowledge store, in the rough manner of a "blackboard
architecture"

-- the possibility to have your algorithm act on data imported into the
Atomspace from NLP, from a virtual world, etc.

-- the possibility to have yoru algorithm help control an agent in a
virtual world (or eventually a robot...), using "glue code" that's already
written

...

The OpenCog framework is basically well-architected but still rough around
the edges in many ways....   It's not as slick as we'd like it to be,
yet....  But for many purposes, I believe it has a lot of value...

-- Ben G



-------------------------------------------
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