I had already found NAO/etc long ago.
Ok, now, seriously, straight-forward, I present my complete compile of how Language works. It's amazing. Paragraph #1 just below helps understand it.
Stored senses and their linked actions, and any linked sets but lesser (ex. clip plays and each frame passes Attention), are only used; sensed (remembered) and acted, when they are selected, by either external or faint internal input's match, self ignition, being saved, or being linked to one. Thinking/Thoughts is internal input and you trying to select stored senses/internal actions to better select and create new interests. Thinking/reading/viewingOutdoors and understanding it without initiating "+" actions does stuff>saves&links senses ex. negatives, affects future searches, fulfills selections, affect action set selections.
> All we select&say about the universe and all we hear&understand is learned by rewards from an input/output environment lying in sets/senses/pattern-recognizers to detect words, sentences, and topic by (like CNNs) totalling up which values heard/exist in network (letters/words/sentences) to select the correct neuron and link to the correct visual image linking to or directly selecting 1 keyword/pre-saved sentence/phoneme and link to actions, and Reversing the pattern-recognizer so selecting senses
(ex. we usually select parameters&close-links/senses&far-links/and sets by internal matches or self-ignitions ex. an image linking to or directly selecting each ex. a Couple key-word/pre-saved sentences/phonemes of which selections could link to more - of which hearing sentences [input] will be understood/heard by the sounds OR auto-Creation of a scene as-hear and its raytrace render both select actually 1 image linking to or directly selecting 1 key-word/pre-saved sentence/phoneme which could link to more)
causes valued words
(value could change as Hear ex. selecting actions/senses make others more likely selected ex. "by" and "past" plus 1 ex. "by" being already stronger or made temporarily stronger i.e. by what heard so far)
(values govern it Outputting something sequencial not "please jar open")
(even motor skill actions since Language Output is actions, even images/videos/information since actions are selected by parameters&close-links/senses&far-links/and sets being selected plus since actions and senses become selections plus since Language Input understood is senses but just the Reverse)
within networks to output sentences
(indeed they flow like water out our mouths/internal-input-tracts, and usually brand-new sentences/words/meanings, and always phonemes/words/sentences that make sense to say back and make sense together, and pattern-recognizer close-links linking "run" to "away from" and "by" and "past" and selecting only 1 for each word you will say).
You can Hear "bad guy on street", recognize, Reverse pattern-recognizer to Say "I'll run by him with a gun", and act actions.
Just some brainstorming for yous:
Does anyone here want to team-up and use our different skills to create AI?
I could, 1 by 1, tell you impressive instructions on what to ut-most first code.
How exciting is that!? You would quickly understand each and see results.
Some of those results are the circuit operating itself, even the way it looks.