Dec 28, 2019, 9:09 AM
Consciousness is deterministic with scale, and biased by body form. Any being with sufficient neural capacity will develop similar intuitions and behaviors. Consciousness means modeling recursive future states by grammar of action. And we are programmed by action in reality.
There is no mystery to consciousness.
The nervous system is a lovely, profoundly simple thing, and there is trivial difference between a transistor and a soma, and the organization (adaptation) of the nervous system by dendritic wayfinding is indifferent from wiring circuit boards other than the infinite untraceable unmeasurable complexity of any given set of excitements. I am pretty confident of the structure of each major region of the brain and how it processes information, and there is no magic to recursive memory by thalamic preservation of state, and hippocampal production of indexes, and hippocampal rehearsal to store memories during break, rests, sleep. The experience of awareness is just a concert, like the experience of consciousness and it’s all just a constructed spectrum of preservation of networks.
And anyone who has written an A* search for way-finding a maze can simply relabel variables and functions with brain regions and grasp the concept. Neural networks approximate that complexity, functions abstract that complexity, classes abstract that complexity further, but the underlying process is the same. Software must make address connections between simulated neurons and serially process calculations, while the brain only makes physical dendritic connections in infinite, inexhaustible, complexity and permutation. At present I dont know (maybe no one does) the provisioning of new episodic indices, and whether they are permanent or temporary.
And the reason we can only produce robots at present is simply one of scale given the problem of serialization. I suspect we will solve that problem as evolution solved vision, which is to break the problem into pieces producing a symbolic input to a hierarchical system of prediction. I don’t see why we can’t do that other than the current cost would be ridiculous although it’s clear google is trying to use pre-processed information (search content) the way the financial business uses pre-processed information ( prices). I don’t know what they’re doing but as far as I know, it requires the development of symmetries (think, metaphysics) and a grammar (think thought, emotion, action) and a vocabulary to do that for the same reason humans needed language to think more deeply (system of measurement).
Right now all I see is categorization (identification), and prediction and I don’t see anyone producing a metaphysics, grammar, and vocabulary bounded by similar rules to language, which will be necessary for machines to communicate because shared hierarchical memory would be slow (other than for the predator-prey division of labor), and protocols (languages) are necessary to encapsulate patterns such that access to (chaotic) underlying memory structures (neural networks and the training that got them to a given state) is unnecessary.