ROBOT1
The starfish learns to crawl like mine did, except it uses a sequence of moves learned which I can set up to learn too. The reason I didn't is because it's the same thing, great, it crawls, but what I did is took the next step, adding why it learns to turn from walls, which is used for all tasks too. Also the starfish may have learned circle actions since it got acceleration and they avoided it! I know how to de-increment the actions! >Dizziness reward, and while it returns to guessing, the strength of the good cue and bad cue both weaken in time to retry!
ROBOT2
Like said it's programmed the pour-drink actions+"idk" and cues. Now, it generalizes by the input barely matching the memory of a similar object, my AI does too, and like the Google Arms it has preinstalled cues and/or actions to-areas and they are tweaked until it grasps the object emerging human-like-grasps. Mine does not! Mine would see a object, save guessed/tweaked head-turn actions that got the reward for object-visual-difference/face/eyes, and again by guesses/tweaks or if told already learned things - it'd then try to grasp. And mine would develop any cue and any actions, not preinstalled *just ball and attempt at-area.
ROBOT3
I believe it has 4 way preinstalled head turn actions to when detects preinstalled visual and/or flicker cues plus tweak hand actions at-target, where mine would guess, then rank guess when seen = learns actions+cue. While change in images add attentive energy for this input to search memory. Now, any nods are preinstalled, blinks too. It may link things? But mine does with all senses and actions (accounting for my attention system) and mine makes (in the episodic mem bar) sets of actions&just noticed senses, AFTer having searched the assorted Primary & Associative Cortex.
ROBOT4
Although the feet motors are limited and I think only guessed in certain directions it still didn't learn to walk really, nevertheless mine and theirs would, it's same as ROBOT1 above, crawl/sequence of actions OK.
So yes, my AI Robot LEARNS to not just mova, but turn before reaches walls, making it the fuuurthest learnin Machine Learning beast. Don't forget the guessing, tweaking as crawls, laughing & crying, and learning to mimic better the reaaal way it had done in shown video.
Face it, I know why we love music and games and knowledge and movies, the neutral non-ranked sense is linked to rewarded senses like good sound linked to french-fry taste and then good linked to kill all men vision, these are mini-rewards, and are super important.