Thanks John.   I am working now on another paper that makes clearer
how the Patterns of Cognition material fits into an overall theory and
practice of general intelligence.

The paragraph you highlight is a critical one, yeah.   Just as the
hierarchical structures in current deep NNs are only approximately
what you want for modeling patterns in physical reality -- yet are
still quite useful -- similarly metagraph folds/unfolds are only
approximately what you want for modeling abstractions needed for
practical transfer learning and generalization -- yet are still quite
useful...   General cognition is just going to be super-slow and
cumbersome by nature, so practical AGI systems need to involve
subsystems operating on varying levels of generality, appropriately
optimized in ways that exploit their limitations...


On Thu, Mar 4, 2021 at 2:15 PM John Rose <[email protected]> wrote:
>
> Anyone who has worked on large complex software engineering projects, 
> especially AGI which tops them all knows about redesigns/rewrites. It's a 
> good sign actually. And it's a form of RSI.
>
> The way to grok this paper, the way I approach it is to scan it back and 
> forth many times then imagine the graphs in your mind, visualize the 
> hypergraphs, metagraphs, operations, inference, learning, mining, attention, 
> then folding and then Galois Connections.  I don't fully comprehend all 
> aspects but I know  Ben has unique perspective on designing and building 
> these things so I at least make the effort.  Hijacking the message thread to 
> ask why it ain't it done yet is really being a troll, is annoying and 
> disrespectful IMO.
>
> This here I find interesting, it's a decision based on much experience 
> apparently:
>
> "Because AGI systems necessarily involve dynamic updating of the knowledge 
> base
> on which cognitive algorithms are acting, in the course of the cognitive 
> algorithm’s
> activity, there is an unavoidable heuristic aspect to the application of the 
> theory given
> here to real AGI systems. The equivalence of a recursively defined DDS on a 
> metagraph
> to a folding and unfolding process across that metagraph only holds 
> rigorously if one
> assumes the metagraph is not changing during the folding - which will not 
> generally be
> the case. What needs to happen in practice, I suggest, is that the folding 
> and unfolding
> happen and they do change the metagraph, and one then has a complex 
> self-organizing /
> self-modifying system that is only moderately well approximated by the 
> idealized case
> directly addressed by the theory presented here."
>
> There's a tradeoff between fidelity and implementable practicality it seems. 
> Go for the low hanging fruit.
>
> John
>
> Artificial General Intelligence List / AGI / see discussions + participants + 
> delivery options Permalink



-- 
Ben Goertzel, PhD
http://goertzel.org

“He not busy being born is busy dying" -- Bob Dylan

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Ta5ed5d0d0e4de96d-M70e7b27d3b67ffac51956170
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to