Any level of complexity creates a natural wall behind it, where the previous level cannot be constructed from the same concepts as the current level; this is just a law of reality. Ergo, any Human devised construct like language, cannot be used to create a true Human level AGI. Language, set theory, graph theory have no inherent intelligence, without the intelligence of the user they are totally useless, they are all a product of an intelligent system, a higher level of complexity.
IMHO the lowest level of abstraction to consider any AGI schema is the spike, electrochemical synapse and neuron. We are the only highly intelligent species we currently know of, it just makes sense to use our own nervous system as template to create an AGI. I needed to simulate the whole system including dimensional space/ time, electrochemical effects, etc in order to get insights into how the brain functions. I think it’s important to draw some distinctions. The brain doesn’t use our traditional concepts of data storage or computation, so no binary gates, numbers, etc. Neither is the brain built complete like a traditional mechanism, it grows and develops along with its senses over time. There are no programs just the GTP, though I suppose for the purposes of this explanation the GTP could be considered as the main program loop. The GTP never stops, it’s constantly cycling and phasing. The brain has a relevant 3D volume/ shape unlike a silicone wafer, so the brain exists in both space and time, which is important. Electrochemical local/ regional effects do contribute the processing as do both the internal and external feedback loops. You can consider reality as part of the feedback process, motor cortex/ cerebellum move muscles, alter reality, vision/ audio and tactile senses bring the loop back into the system, etc. Memories are engrained in the structure of the connectome, relative to the current state of the GTP so no linear/ serial searching is required, everything’s parallel. There is no inherent learning mechanism, the connectome learns to learn, it’s a structure that has the capacity to learn anything within the frequency domains of its senses. The external sensory cortex re-encodes incoming sensory streams by applying spatiotemporal compression and then re-maps the temporal phase to a spatial dimension/ location, time is irrelevant to the mapping. So the external sensory cortex areas recognise/ encode to external reality time, but once the sparse re-encoded signals enter the GTP they are outside/ reality time independent, internal GTP time is regulated/ utilized by the connectome. Our sense of time is provided by another brain structure entirely and can differ greatly from the external/ reality time, the brain can change its concept of time depending on the experience or state of the GTP. The GTP flows through the intelligent structure of the connectome, and the connectome is built from the state of the GTP. Curiosity partly comes from the sharing of common networks, cortex areas that encode similar qualities of reality. Self-aweness is the recognition of internal processes in the same terms that the system understands external processes. Consciousness is a bi-product of the complexity, harmonics piggy back on the logical patterns within the GTP, it can’t be avoided and is learnt and engrained into the connectome along with everything else. If the GTP stalls/ stops then that connectome is dead. I’m ranting, I’ve been doing this project for 20 odd years and I’m still awe of the complexity and beauty of the human condition. > Which sounds like you want to replace phonemes with some kind of protocol based on "what it has heard from its peers". The vocal output streams are not directly related to what the system has previously heard. They are an internal approximation built from the various cortex regions that have contributed to the output. They are generated by the systems internal model/ understanding of reality through experience using its own senses. Phoneme model I was using the CMU dataset to test the audio clips of the phonemes I’d recorded from the phoneme video, I needed to be sure that the clips would in-fact when combined produce intelligible speech. The actual vocal outputs from the connectome are parallel spatiotemporal spike trains that would normally drive the vocal cords/ muscle inflections used to shape the sounds, so the pronunciation of a phoneme is actually a stream not a specific trigger. You can feel your jaw, tongue move as you make sounds, you can hear the feedback through the bone and these signals are all relevant to the connectome, they drive/ provide the resolution/ accuracy required to produce speech. At this stage the phonemes will require a kind of wrapper; a program to interpret the connectomes output and then trigger the phonemes. This is not an ideal solution hence the botch, but I’m already using a similar schema to convert the bots head/ arm movements. To me it’s not a matter of writing a theory; we already know of an intelligent schema, we just have to figure out how it actually functions and build it. There is more information @ the following... https://www.youtube.com/user/korrelan https://sites.google.com/view/korrtecx :) ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf97c751029c2e4db-M6238e7e5cd342c968571ee19 Delivery options: https://agi.topicbox.com/groups/agi/subscription
