[goaty]

Your system looks good,   but im only seeing you finishing this thing in the flesh/plastic whatever if you can make these brain bells in batches.

Because you've got 1k to make!   so there needs to be some manufacturing wizardry to get it done.

[HS]

Yeah... Maybe some big silicone ice cube trays, and reverse mold those with just magnets, to then get automatic magnet placement on the inside.

[goaty]

Just takes more brainstorming and then you might find yourself with your own technology company out of your basement... try not to go mad in the process,  if u end up being able to produce them for 5c each, it may send you around the twist, or filthy rich, or both.

[Korrelan]

You could make furniture out of them, I'm just going to sit in... and talk to... my thinking chair lol.

 

[HS]

Yes... yes.. might see some cool effects if I incorporated  memory foam into the design aswell.

[goaty]

memory foam?
 

[HS]

It's one of those things developed by NASA for space, but ended up being so good that it's being used on earth for everything, furniture, etc, like duct tape.

[HS]

I've been struggling with how to assemble the artificial neurons efficiently. I think the answer is a 3D printed framework to hold the magnets in place. Three additional improvements are also made possible with this design; making the magnet pockets 2 magnet thicknesses deep will prevent lateral slippage of inter-cubic magnets, the opaque plastic will provide distinct windows for better led signaling, and it will also allow me to chamfer the edges and corners as they were wasted material and also quite spiky.



Of course I'll still add the resin so it's not janky.

[Korrelan]

Nice, though I think you might have a problem accessing it for soldering, etc.

Have you considered a negative mold, where you can split it apart (ABCD) to enable easy access for soldering the components? It could also hold the connections in place until the resin (blue) is poured.

https://photos.app.goo.gl/UDRVoYpdpfLZtNjK6

 

[HS]

It might be enough to print the bottom 2/3 and top 1/3 separately, and just have the top be a lid. But if it isn't, I'll fall back on your idea.

[HS]

So, I've got the mental index of narrative intelligence to expand spell out in greater detail:

Goal

[   
fight [],
flight [],
hide [move to darkness, become dark and still],
bask [soak up sunlight],
play [attempt uncertainties],
Sleep [e-zzz, I should hope]
]

Conflict

[
uses reflexive Intelligence [],
events [Surprises [new actions, and new reactions to actions]] to memory[],
processed [] in sleep
]

Result (apropos success)

[
yes, (contentment),
no, (interest and fear),
yes, but new problem, (interest and contentment),
no, and new problem, (fear),
]

Emotion

[
interest [] [ comparison to opportunities[]],
fear [] [ comparison to dangers[]],
contentment [] [ random association[]]
]

Reason
[
using comparisons [] and arithmetic []
]

Anticipation

takes result of reason and matches with best precursor to choices

Choice

[
fight [preparation []]
flight [preparation []]
hide [preparation []]
bask [preparation []]
play [preparation []]
sleep [preparation []]
]


I'm also narrowing in on a good and easily manufacturable artificial neuron design:



It's just one part, flipped over and turned 90 degrees, for bulk printing pricing advantages. It's got some neat features, let me show you! Each face has four arrow shaped windows, allowing LED's to indicate what the cell is doing and where the signals are going. The holes for the wires (in the four corner magnet sockets, on each face of the cube) are positioned so that the magnets will end up tilted away from the center, reducing the risk of any crossed wires due to spring contact. Those same magnet sockets also have gaps to allow for the electrical connection to the other corner magnets, greatly reducing the headache of internal wiring.

The big question is how on earth to bridge the gap between the mental index and the artificial neurons. But The Mental Index is TMI for a single cell. So perhaps each cell will have a specialization, or activate a stored specialization from the index, based on if it's the 1st 2nd 3rd etc, to process a signal.

[WriterOfMinds]

I think that narrative intelligence loop is my favorite of the ideas you've introduced to the forum.  I'm interested to see how it plays out when you get it implemented in a system.

[HS]

Well now it seems obvious. This seems a rational way to organize a brain inspired structure. The ironic problem of many artificial intelligence theories could be that by trying fill up the head with too many brains, they left insufficient room for systems which are supposed to run in tandem with pure thinking in order to prop it up.



The entire structure of this cross section would be composed of 144 cubes! I could go one size smaller like so:



But I think you've got to have several of each type of neuron to develop some reliable basic functions for this theoretical system. Not to mention it'll be cooler, which should or course be the ultimate deciding factor for any Ai dream. 

I'll need three unique codes, inciting three types of neurons to play off each other, to hopefully get this system churning. Input goes in through the filter/bouncer created by the autonomic cells. Simple signals get bounced back, allowing for quick reactions to events interacting directly with the body. Complex signals get kicked up the ladder, into the next concentric dome of neurons. Then an appropriate emotional response can be estimated, then in the next dome events can be planned to align with the will of the emotional layer, altering it in the process.

Maybe inputs will generate an emotional opinion/consensus when moving through the limbic layer, and then the cerebral layer could adjust the emotional response with the E part of its GCRERAC process. This would imitate the top down effect of rumination or positive thinking; allowing the cerebral layer to modify the brain's prevailing mood as one of its goals. Thus creating a sort of reverb optimization which seems promising in terms of creating a self sustaining process, and maybe even eventual consciousness.

[krayvonk]

Looks great.  That narrative intelligence thing looks promising too.
Wouldnt it be easier and youd get more to mould them instead of printing them?
:: Maybe u could get the mould printed - then batch mould it yourself.

Designing and making your own robot from scratch ends up a lifes work,  it never ends.

I have something interesting to show u about milk -> 

[HS]

From a first principles stand point, yes. But this would require a larger upfront time and money investment. I would need to design and pay for the manufacture of a custom, probably three-piece silicone mold like three ice cube trays fitted together, for a design which probably still has flaws to be ironed out. I'm guessing it will be best to do 3D printing for mid scale production (hundreds), and molds for large scale manufacture (thousands). The milk thing is cool, thanks!