Ai Dreams Forum
Software & Hardware => General Hardware Talk => Topic started by: ranch vermin on October 06, 2017, 12:19:52 am
-
Something interesting if you wanted to make a really small robot by yourself if your a good electronics human!
This is delay line memory, it was done alot in the 1950's and it gives you memory without having to put the bits down physically, and a computer can be made with this with not much more than a handful of parts and requires no minification, except managing the delay offset, and inserting the pulses into to repeat.
https://www.youtube.com/watch?v=-qerYLM-eEg
-
a computer utilizing the memory can look sorta like this->
(https://ub.fnwi.uva.nl/computermuseum/pictures/p101dl.gif)
the parts in the middle are the gate, and around the outside is the memory delay, where the bits live temporarily.
But probably this actually isnt a full computer, but im pretty sure its about this big - even for a full computer, possibly with a fixed program only.
-
I like delay lines. I know they also made delay line with mercury filled tube.
I also believe the brain uses delay line of neurons daisy chained together. A spike traveling around in two or more neurons
in a loop:-)
-
Quite possibly, that there could 2 or more pulses in a synapse fire? who the hell knows.
-
I have a AGI model that uses daisy chains of spiked NN. There is the traveling pulse that jump to the next neuron. And
then there is second pulse that goes into a rather large spiked NN. In this case the output from this Large generative
nn is video.
There are endless lengths of daisy chain in the brain. This system is paired with a detector CNN. When the CNN
detector a cat a pule is sent to all the location in the daisies chain where cat is.
If a traveling pulse in the daisy chain is hit by the cnn then this will re enforced it. Other wise it will fade out.
If a CNN pulse goes to another location then new traveling pulse will start and cause the video to jump to that location.
-
this pulse spike jumping from one chain neuron is a focus pointer in a organic fashion.
-
that sounds interesting, sounds like what a boltzmann machine would do.
Mines similar as well! But i learn everything out the corners of the screen, then i join them together to remember the thing, to play it back. (output of mine is video too.)
It should be able to return the backsides of objects too, by connecting corner to corner.
And it can generate 3d, by comparing different samples of the same corners together.
-
My NN agi brain paired with a chaining NN. Iitcan do amazing things.
The generator nn and daisy chain of neurons are to
keep in sync with reality of time. The chaining NN take in data and transforms to output data. That is feed back into
the front of the the chaining NN. So that it generates of list of thing to do or a list of what is expected show up in the video.
Like a link list or a one way hashing.
Chaining Neural network gets it source info is form output of it's self or from video or both at the same time and can be
small fraction of info form the source. What every the is being FOCUSED upon. A small fraction of the video can be
a cat extracted from a video frame or the difference error of what was expected.
-
cool.
imagine making movies with one! with faders and checkboxes, not just controlling a character anymore, the whole scene!
-
AI Learns To Recreate Computer Games | Two Minute Papers #195:
https://www.youtube.com/watch?v=2VyhmbEjs9A