Relevant threads
- By virtue of reflexive and reflective reasoning then reflexive and reflective thinking?
- FROM SIMPLE ASSOCIATIONS TO SYSTEMATIC REASONING: A Connectionist representation of rules, variables, and dynamic bindings using temporal synchrony
- The paradigm shifts in our minds
- Handles-tags drag along thoughts from our minds
- "Use-it-or-lose-it" pruning of brain cells. An on-the-fly brain mechanism for dynamic minds?
I read in the paper by Lokendra Shastri and Venkat Ajjanagadde 'From simple associations to systematic reasoning', under the title 'Computational constraints', section 1.2
"Connectionist models (Feldman & Ballard 1982; Rumelhart & McClelland 1986) are intended to emulate the information processing characteristics of the brain — albeit at an abstract computational level — and reflect its strengths and weaknesses. Typically, a node in a connectionist network corresponds to an idealized neuron, and a link corresponds to an idealized synaptic connection.
Let us enumerate some core computational features of connectionist models: i) Nodes compute very simple functions of their inputs, ii) They can only hold limited state information — while a node may maintain a scalar ‘potential’, it cannot store and selectively manipulate bit strings. iii) Node outputs do not have sufficient resolution to encode symbolic names or pointers. iv) There is no central controller that instructs individual nodes to perform specific operations at each step of processing."
A node, in a connectionist or any other kind of network for that matter, corresponds to an idealised neuron, and a link to an idealised synaptic connection? Taking it further, a node corresponds to neural pathways connecting links, pathways which can include one idealised neuron or several idealised neurons? The main point being the passage of the signal from link-to-link? And not just any link, but a certain kind of link? A link, a synaptic connection, that lies within the context, the link-node-link path is attached to? A neurological basis of context? That links are attached by nodes, in link-node-link neural chains abiding to rules (... of context?)
Link-node-link creation, independent of length, namely the number of neurons involved. Independent also of the distance between individual links, or their place in the brain? What only matters is their placement within the contextual link-node-link chains? And one other thought that connects our brain organisation, with chaotic and fractal aspects. It is the potential inherent in such mode of brain organisation that even a remote, unused link, deeply buried, in all sense implied, can instantiate itself, unearth the attributes is attached with, and confer in the meaning of the currently active link-node-link unit.
Take the matter of idealised units, beyond a computational level, and instead talk about idealised units on the basis of context? Providing a simple mechanism for instantiating a thought, a unit of emergent thought, and multiple copies out of the same blueprint interacting, leading to complexity and the emergence of the mind?
The little red riding hood example and the steps in the inferential processing,
"The wolf will approach LRRH (to eat something you have to be near it); LRRH will scream (because a child is scared by an approaching wild animal); upon hearing the scream the wood-cutters will know that a child is in danger (because a child’s screaming suggests that it is in danger); the wood-cutters will go to the child (people want to protect children in danger and in part, this involves determining the source of the danger); the wood-cutters will try to prevent the wolf from attacking LRRH (people want to protect children); in doing so the wood-cutters may hurt the wolf (preventing an animal from attacking may involve physical force ...); so the wolf decides to wait (because an animal does not want to get hurt)."
a link-node-link path, and each link jump adds attributes that amass in the meaning conferred by the link-node-link chain, in a manner that comes out, from what the authors refer to as the “unary or even propositional fixation” problem
"This turns out to be a difficult problem for neurally motivated models. As McCarthy (1988) observed most connectionist systems suffer from the “unary or even propositional fixation” with their representational power restricted to unary predicates applied to a fixed object. Fodor and Pylyshyn (1988) have even questioned the ability of connectionist networks to embody systematicity and compositionality."
as 'unary predicates applied to a fixed object'?
Links thought off as handles or tags that involve their own specific load of attributes and properties, and drag along what is relevant to the context of the active node-link-node chain?
Though their potential surpasses their mere role in providing inferences for instantiating reasoning. By virtue of their associations with other conceived thoughts, that might even belong in other contextual units, can form context to context bridges, can drag along attributes that belong in other contexts, borrowing from context to context, passing from context to context and use that in innovating manners, the hallmark of creativity.
