[krayvonk]

The milk thing is cool, thanks!

its 5 times cheaper than epoxy resin! =)    probably not as strong, and theres more tricks to learn about it,   one of the extra tricks is soaking it with cold tea, for a short time, and it improves the strength.   takes about 2 days to dry,   I think you can put it in the microwave and it dries quicker.

What im planning on doing with it,  isnt the brain (which is a very difficult thing to attempt....) just making motor-servo joins and brackets and the main framework for the robot, resin seems like a really good idea,  you wont need a single screw to do it that way.  (its not as tough,  bolt together stuff is tough as nuts and bolts.)
That job is fairly simple in general, the body of a bot, everyone can pretty much handle it, its just like lego after all, and resin is a cool way to do it.

[frankinstien]

You've got an interesting idea. 

Have you thought of placing those cubes in a colloid, populate it with varying cubes with different contact points and let them swim around to find each other? You could use an electro-magnet(s) in the cubes and power them with a microwave or RF signal externally. Each cube could then buildup resistance to inputs depending on how many cubes end up connecting to it.

An alternative is to use LEDs and photodetectors as the actual communications between cubes and just use the electromagnets to move the cubes about in the colloid. Now here's the real interesting part of using LEDs for communications: You could then communicate with the formed circuits of your cubes through the LEDs and photosensors as well! So these cubes float in this clear gel and you could form say a sphere with LEDs and photosensors around the container of the colloid with cubes and now you got a means of 3 dimensional parallel I/O from the circuits formed by the cubes. 
 

[HS]

Thanks, that's where I would like this technology to go. Vast numbers of artificial neurons suspended in a stable, clear, heat conductive gel like substance. Tiny bots could then navigate around the brain clearing out damaged neurons and replacing them. I like the idea of it being a colloidal mixture, it would reduce the need for precise density matching between the neurons and the medium they are suspended in. The tricky part is getting the programming/types of neurons dialed in enough that once this mixture is injected into an appropriately equipped cranial cavity, they would self organize themselves into a generally intelligent network.

The apparent options for non contact communication such as light, sound, radio, or Bluetooth, as well as the miniaturization required, proved too difficult and/or expensive for the time being. So, I've left that part of the system sit until inspiration strikes, or new technology becomes available and affordable. I've made the communication physical by using electrical connections while I think about the programming aspect of it. At first in broad concepts, but hopefully narrowing down to the specifics which could produce useful results. 

You've seen my attempt at mind molecules in that thread with the picture of different types of neurons fitted together, where I'd hoped these mind molecules could be fitted together like a 3D Tetris or a homogeneous solid of some kind. The closest I've gotten to that is a simple crosshatch which generates multiple 3D networks within each other.  Neurons remain in contact with some types more than others, but they still end up enmeshed in a better approximation of the random distribution which would occur in a gel, while still allowing neurons of the same type to be in direct communication with each other, like what would also be possible in a gel.

Light as a power source seems to be the best option. As for communication, wireless waves of unique frequencies for the emitters, and individually adjustable frequency filters for the receivers look like they could create a tunable network.

Eventually, by having something like a field programmable gate array in each physical neuron, it might be possible to have an adjustable net governing the function of each neuron. That way the network could become more compact by reducing the total number of required physical neurons, thereby increasing the total computational capacity per set number of nodes, and become capable of fine-tuning itself to a greater degree.

[HS]

I'm switching the design, the Blinky robot will consist of just one cube. That will bring it closer to the original premise of a tiny aquatic sun feeding creature, it will simplify the programming, and I'll finally have a legitimate reason to print out and assemble a couple of these! Blinky will still run a narrative loop, but it will be self contained instead of a huge communication network. I'll still try to make multiple cubes and allow them to interact, but they will be individuals rather than a collective system. If I'm able to do that, then for my next project I might try to make something multicellular.



There are jet tubes on the edges, water would be pushed through by internal propellers. The blue circles are solar panels. The arrow shaped windows are for signaling/communication/sensing. Electromagnets on the springs for docking/holding onto things. The two halves of the shell are slightly offset to show how they fit together.

[HS]

Woops. Project creep, I guess. At least now it should be able to move around a bit faster. If Blinky turns out to be sea/sky worthy, UFO sightings will definitely increase in my area.
 


So, what's new? Well I got inspired by the spherical Westworld brain design. That should help with creating a reliable center of gravity, all the batteries and processors will go in there, it will also glow different colors for communication. This pairs nicely with a cubic interface, that way you get three layers; sensors, cables, and data handling. I increased the rotors to a more reasonable size, and subtracted electromagnets in favor of quad copter cameras, (which will go in the center holes). The idea is that the Blinky's will observe and equalize each other's color temperatures as way of communication. Blue would indicate high alertness, and a red would communicate relaxation, these would alter the thresholds for certain behaviors.

[HS]

I got the 30-day trial of this program, so there’s a bit of remodeling happening here. If you should spot any possible problems, improvements, or have any suggestions, by all means let me know, I still have plenty of time continue inventing, and to make adjustments.

Attempting to simplify Blinky into one brain cell may have been an oversimplification. I’ll be trying to apply some of the concepts from ‘Human Style AGI’ to the Blinky project. Right now, a design of 27 neuron cubes serving as Blinky’s brain seems like a good place to start. I’ve decided to simplify the eyes instead of the brain, using solar cells for both charging and sensing. Some edges have intentionally been left sharp and angular, because this keeps the part file size low, my computer is barely able to handle the main assembly as it is. The solution to that would be to merge/simplify more parts, which would mean a loss of properties/functionality, but at least the program won’t crash.

The cubes (2 x 2 in, 5 x 5 cm) are now hollow, mainly to avoid explosions in case a battery overheats, but a little weight reduction, and shock absorption resulted as a bonus too. I’ll still have to figure out a way to increase the heat transfer, but at least no explosions.

Speaking of which, virtual heat transfer might not be a bad starting principle for neural signaling and motor actuation. Same principle as the real thing, only sped up. Huh, another thought, if only I could make the signals analogue… It might allow a nervous system to fine tune itself to godlike levels of precision! In Blinky’s case it would only make him godlike at finding sunny spots. But still, it’s a thought.




The rubber tubes at the corners (seen in the image below) holding the electronics net (the black mesh with glowing red, green, blue, and white cubes) in the middle of each neuron cube, can vent exploded battery (just in case…), and can later double (with the addition of wires) as antennas, if that should become helpful or necessary.

I think I finally managed to imagine a good way of securing the springs to the magnets! As you can see in this cross-section, there are two magnets on each end of a spring. Since the thicker magnet is countersunk, that shape will hold the spring in place once I pour in some solder or resin. Ideally, I’d like to use solder, just to be extra sure everything is strong, resilient, and electrically connected.

But the hot metal may reduce magnetization, and could mess with the tempering of the spring. Instead, I could apply some conductive glue to the end of the spring, stick it to the back magnet, then carefully pour in some resin.




Most of the process looks like this, and involves solving little problems such as how to keep this ring magnet from detaching and falling in (instead of falling out) which was the previous issue.



If I do the obvious thing and embed it halfway in the neuron shell, how am I going to attach the parallel power distribution net? (That’s the red and blue rings and leaders on the inside of the cube.) Will I have to redesign the power net? Ok, but then it might prove difficult to manufacture... Eventually I’ll be walking my dog or something and suddenly exclaim ‘Eureka!’

The four, colored, glowing cubes (they represent 4 parts, but I temporarily added 8 for symmetry, to help in the assembly process) at the center of each artificial neuron, represent: Red - Signal Receiver, Green - Signal Emitter, Blue - Signal Modulator/Memorizer, White - Battery. I will need to research what’s available, and will probably add more components later.

A look at the parallel power distribution net: The idea appears to be working, and scalable! The neat thing about the power distribution net is, it allows you to connect the cubes in any orientation. By the amazingness of geometry, (provided there is a single, or multiple non-contradictory power inputs), intact neuron cubes won’t be able to short circuit.



The interior case, holding all the artificial neurons, will need ports for: Charging the batteries, individualized light detection from 6 solar cells, and individualized power outputs going to 12 motors.

I’d like to do all this without the need for external electronics. Meaning that no components will be required besides what is already in the neurons. Ideally, the robot will only need the solar cells, the outer case, the inner case, the neural net, and the motors/propellers.

With the exterior springs attached, and half of the interior case on: The completed internal case is a double of the same part, only flipped and rotated 90 degrees. The center of each face has 4 ports, which must somehow fulfill the three purposes of, charging, light detection signal conveyance, and motor actuation. The barely visible magnets imbedded in the corners are meant to hold the two halves together.



A look at the motor housing, on a piece of the exterior shell: This will be composed of 8 identical pieces. The groove on the bottom right fin is for wires going to the motor.




1/8th of outer shell with motors with propellers, plus how they will attach to the neural net case: The props are designed to work equally well clockwise and counter-clockwise. Again, the barely visible magnets (on the edge of the motor housing), should help to hold the outer shell together. I was also trying out some environment options, but didn't find an ocean environment. 'Old Warehouse' seemed like the next best thing.




A cross section view of the entire system: Everything fits together nicely! Except, perhaps the inner shell could be a little thinner for weight reasons. Blinky is meant to be aquatic, and sun feeding, so I’m aiming for slightly positive buoyancy.



The initial concept is completed!



Well there it is. The rough outline of the vessel which, I hope, will eventually be capable of containing, expressing, and sustaining, the coursing energies which will be Blinky. One day, little buddy, one day…

[MagnusWootton]

The design is cool, very professional,  but I think 27 neurons is a bit skimpy,  even if you made them analogue (which i recommend) it still would only be 27 adds parallel or series from input to output,  if you had 10x10x10,   1000 adders parallel or series, thats more what would work in my opinion.   (to contain the truth of its environment.)

[HS]

Thanks. As you say, the final version of Blinky may work best on about a 1000 neurons. When I’m able to create multiple neurons cheaply, I’ll definitely be increasing the amount. The idea goes like this. Design the actual resin parts, design the rubber moulds for casting the resin parts, design the parts for creating the rubber moulds, get the parts for creating the rubber moulds 3D printed. Then I’ll be able to create, (and continually increase my capacity for producing), the complex shapes of the actual robot, by using only liquids!

For now, I’m hoping that wireless communication between the 27 neurons can increase the interconnectedness of the net, and simulate the effects of a larger number of neurons. The Blinky project, and the mobile-wireless-artificial-neurons for my eventual AGI / Human repair project, seem to be converging into one design. You could say Blinky is a huge scale artificial neuron, only with different programming. So eventually, I’ll definitely be taking this to the max. Billions and billions!! That may also mean waiting for price reductions in current electronics / the development of new types of electronics. But that's far in the possible future.

[MagnusWootton]

The idea goes like this. Design the actual resin parts, design the rubber moulds for casting the resin parts, design the parts for creating the rubber moulds, get the parts for creating the rubber moulds 3D printed. Then I%u2019ll be able to create, (and continually increase my capacity for producing), the complex shapes of the actual robot, by using only liquids!

Casting is the way to go for large amounts of things,   if you cycle things as quick as possible, and do as much as u can together in biggest batches you can, it beats 3d printing by a huge margin.  then instead of struggling to have 1 blinky, you could exceed this by accident and have an army of blinkys! hehe.
It may seem daunting to make a 100 cell system or something at first,  but if u keep brainstorming how your putting it together, it might get easier if u come up with a more efficient production method,    I wouldn't say its impossible to get to even 10k maybe, but it depends on how you do it.   The more you concentrate on one thing, the more perfect you get it, given time.

Thats how we get everything made, by honing production techniques over time,  everything ends up too productive, u end up with boatloads of over-manufacturing cause the technique got honed down too much!

For now, I'm hoping that wireless communication between the 27 neurons can increase the interconnectedness of the net, and simulate the effects of a larger number of neurons.

Definitely agree if you can get the amount of synapses up, without increasing the complexity of the machine using some wireless communication thing, the cells would count for more, because it would get more efficient use of the same space for storing the same method.

Analogue will save you more space too->>

You can get a 1024 logic element FPGA for fairly cheap,  but if you take advantage of analogue, a home made system could be the equivilent of 10 or even 20 times more than that, because they are pure digital they are at a disadvantage when it comes to storing precision.   its *precision times worse under analogue, because analogue gets its precision for cheaper, real estate wise.

its about 16 logic elements for an adder,  but its only 1, if its analogue.

[HS]

My plan seems to have worked! Now I should be able to manufacture complex shapes using only liquids!



Regarding adjustments, I think I'll just get rid of the interior case altogether, and have the power shell and connectors embedded in the exterior case, like I've done with the neurons. The magnet problem was solved by using magnets of half the thickness, sandwiching the power shells between the magnets, and sandwiching the magnets between supports on the rubber moulds.

Now I've got to check if, or how, this theoretical process and resulting device will actually work, make a list of improvements, and probably do some remodeling next year. If it works well, and I'm able to cast the neuron cases cheaply, then I will indeed aim for more than 27.

[MagnusWootton]

What is the liquid you are using?

If your getting shrinkage problems, You can take a few more steps, If you form it and set it layer by layer it gets rid of the shrinking problem 100%.   Because it sets quicker as layers then it actually doesn't take longer to do,  but I guess its a bit more energy.

Heres a guy making a mould (but you can also do the casting like this as well.) out of latex.

[HS]

Silicone and epoxy resin probably. If something doesn't work well enough I'll switch to a different substance, that or change the design to correct for the problem. Most of the fun/work for me will be to figure out how to make the process simple and non finicky, I think that's the only way I'll be able to do it at scale.

[MagnusWootton]

Sound your getting there, looks like your putting the work in!   if you get Blinky finished you could be amazingly rich, im guessing.    Its worth it.
Just imagine how much you can make (I mean other products, not money) if U can make a robot,  a robot is the hardest product to make!

I'm working on my stuff too (I want to get there as well ),  Ill have a post of something I'm doing in  S - CAD,  its quite a good one, probably 1 day from now.  I just need a little rest, then Ill tap the rest in for it.    Its 99% there, just a bit to finish off.

[HS]

I'm trying something new and setting a goal for myself: Before the end of the year, I'd like to build a ground-surviving first iteration of the All-Terrain Blinky illustrated above. Because I won't have much of a budget, and since I somehow managed to make a hinged oscillator component with it last year, I'm planning to make most of this out of cardboard.

This first version of Blinky almost certainly won't be sea-worthy, or powerful enough to fly, but if he can navigate around the house by himself and charge in patches of sun/lamp-light, I'll be more than happy.

[infurl]

That would be really awesome if you could accomplish something like that. I've seen drones that fly and go in the water. I've seen drones that can fly and roll on land. I haven't seen any designs that could do all three though. I have one suggestion; have you thought of experimenting with whegs instead of wheels?