Using this Brain Twin and the largest NMI, the
Brain will be transformed into the largest
machine.
This first NMI was recycled into a much
larger unit. It could attach to the Brain and
the Brain Twin to create the largest Humanoido
Machine and Brain to date,
If (and when) the primary Big Brain is connected to the Twin Brain and connected to the neural matter injector that was originally completed, the size of the Big Brain will exceed the size of the largest Humanoido project built to date, which is the UltraSpark 40, a 320 Propeller Processor machine.
In terms of power, which should be defined first - is that quest to engage (among other things) the maximum amount of neural containment, also classified further as neural substance, neural matter, and neurons.
Note: in our definitions of neurons, we will define a machine neuron and not that which comprises a human neuron. Machine neurons can be manufactured to dilate once injected, thus lowering the numerical requirement for a specific brain containment.
The objective of the Big Brain is not to replicate the details of the Human Brain (unless absolutely necessary for some reason), but rather only use the brain as an outline and reference.
Dilation is an algorithm that will be explored when the point of neural propagation is reached.
In the past, a rotational matrix was used for conditioning containment fields, and so if any budding mathematician programmers want to pave the path for neural matter dilation in a Propeller chips propagation of neural matter, go ahead. A matrix equation could be a future topic.
I am now convinced that machine life can be established with the Brain. In studying the methods and developing methods to lead to this objective, it becomes very important to define exactly what constitutes life, and what comprises its various components.
We are probably not interested in conception and propagation of little brains as procreated by the big brain, though reproduction may be in someone's definition of life. Yet, there are many people in life who, for one reason or another, cannot have children. These people are obviously "alive."
Also as pointed our earlier, if you cannot turn the door knob to open the door, your level of intelligence can still be off the high end scale of IQ.
Yet, it is not often that we have opportunity to meet genius, and when my Physics professor ribbed me hard during the Astrophysics conference in Lincoln NE, I looked up only to discover that one of the world's greatest Astrophysicists of all time, MIT's Professor Morrison, was sitting right next to me!
During that day, he taught things about the Universe that would never be forgotten - but I learned something else, that in his quest of genius level understanding, his body devoted all its physical resources to his brain. This indicates that devoting lots of resources to the machine brain will be important for the highest level of intelligence.
So in the light of the doorknob, it seems relatively unimportant when it comes to a level understanding of gravitational evolution and converse devolution. Yet, based again on a definition, the turning of the doorknob may be a priority.
On to the Key. What contributes to making a machine alive is its given purpose, programming, routines, heuristic algorithms, subroutines, level of programming, architecture and application. In a machine, we can bring forth different levels of life, styles of life, let the machine have leeway to make decisions and learn, set the rules, be inquisitive, and govern the applications and set the domains.
It's about the life of anything that can be understood and defined, and then transformed into the machine. We are the master, but one school of thought is we teach the machine and let it learn beyond its programming. Again some definitions are needed. What is "beyond its programming?" This may be as simple as extending a given database, asking questions that were not specifically programmed, and having some originality or creativity when painting.
Again, as with painting, we are dealing with a program, is it water based, oil based, charcoal, pencil, color, B&W, conceptual materials, etc. and what is the style of painting, i.e. is it like a Picasso, Rembrandt, Monet, da Vinci, Van Gogh, or like a machine?
I suggest we look at different levels of intelligence base on varying applications. Again, we need to define applications in some highly specific ways.
There's a lot that goes into opening a door. First an AI robot has to identify the door and have a reason to open the door. Doors are made out of all kinds of materials and come in many sizes and styles. The robot has to figure out how the door operates; push, pull, slide, revolve. What are the door mechanics; knob, lever, handle, automatic, wheel chair access button. What do you have there... Purpose, observation, recognition, memory, problem solving, maybe some learning if this is a new door experience.
There's a lot that goes into opening a door. First an AI robot has to identify the door and have a reason to open the door. Doors are made out of all kinds of materials and come in many sizes and styles. The robot has to figure out how the door operates; push, pull, slide, revolve. What are the door mechanics; knob, lever, handle, automatic, wheel chair access button. What do you have there... Purpose, observation, recognition, memory, problem solving, maybe some learning if this is a new door experience.
Excellent characterization of handling the opening of a door. I might add the importance of gripping force, having a hand with non-slip material with a decent coefficient of sliding friction, calculating the gripping force so the handle is not crushed or too lightly held, not to mention the strength required to open and close the door, plus checking to see if the door latched properly. If the door is an outside door, a huge wind may be a concern, so it's important to sense the wind force. If it's a circular rotating door at the Mall, there's the extra contention of being careful if someone is in front or behind, and sometimes these doors require pushing and not pushing, and it's a trick of timing to walk and exit at the perfect speed.
While opening a door seems trivial as a human skill, it is a rather advanced skill for a robot.
Mike G and Humanoido make good points about the various challenges for a machine to open a door it (not ready to call them he/she yet) hasn't encountered before.
I believe a lot of thought has gone in to teaching robots to climb stairs and open doors. It seems it's easier to teach a robot to fly than to teach one to travel up a couple of flights of stairs and turn a door knob.
As you may know, while not a primary objective to the "Fill the Brain" project, there is some use in reverse-engineering the human brain, or animal brains - not as a whole but as a fractional part.
The human brain is made up of around 100 billion neurons. These neurons connect and communicate with each other through a massive network of around 100 trillion synapses. This indicates the great complexity required to make a human artificial brain.
There's a trick that the human body uses. Genes don’t map out every single detail - they give a more general instruction and repeat a few million times (i.e. Fractal Instructions).
One such repeating structural unit is called a Neocortical Column (NCC) comprised of a group of about 10,000 neurons in the Cerebral Cortex, organized in a relatively consistent way across a brain. Grey matter contains millions of these columns.
Rather than trying to create a model of the whole brain at once, one idea is to create an artificial column that responds the same way that biological ones do, i.e. built a virtual copy of an NCC and replicate it for real-world activity.
Thus, the current brain crop can handle two Neo Cortical Columns at 20,000. The 1st NCC can be cloned, just as we clone the neurons. Just what we can do with two of these columns remains to be seen.
However, what we can learn and borrow from the NCC is the general repeating function. If similar compacting ratios can be achieved, i.e. one million to one, it could have serious impact and implications on the machine brain.
There's a trick that the human body uses. Genes don’t map out every single detail - they give a more general instruction and repeat a few million times (i.e. Fractal Instructions).
One such repeating structural unit is called a Neocortical Column (NCC) comprised of a group of about 10,000 neurons in the Cerebral Cortex, organized in a relatively consistent way across a brain. Grey matter contains millions of these columns.
Rather than trying to create a model of the whole brain at once, one idea is to create an artificial column that responds the same way that biological ones do, i.e. built a virtual copy of an NCC and replicate it for real-world activity.
"Fractal Instructions?" You're just parroting buzzwords. When you quote or paraphrase someone else's work, you -- at the very least -- need to give credit, viz:
"Fractal Instructions?" You're just parroting buzzwords. When you quote or paraphrase someone else's work, you -- at the very least -- need to give credit, viz:
From the Blue Brain Project article (BBP), cited above: "Genes don’t really code the body like blueprints do for a building, mapping out every single detail; instead, they give a more general instruction and hit the “repeat” button a few million times (e.g. when they give fractal instructions)."
Humanoido: "Genes don’t map out every single detail - they give a more general instruction and repeat a few million times (i.e. Fractal Instructions)."
BBP: "One such structure is called a neocortical column (NCC): a group of about 10,000 neurons in the cerebral cortex that are organized in a relatively consistent way across the mammalian brain. Millions of these columns compose the whole of your grey matter."
Humanoido: "One such repeating structural unit is called a Neocortical Column (NCC) comprised of a group of about 10,000 neurons in the Cerebral Cortex, organized in a relatively consistent way across a brain. Grey matter contains millions of these columns."
BBP: "Rather than trying to create a model of the whole brain at once, the Blue Brain Project is attempting to accurately model a single NCC in a rat’s brain. If they can create an artificial column that responds the same way that biological ones do to electrical impulses, they’ll be on the right track to building a good model."
Humanoido: "Rather than trying to create a model of the whole brain at once, one idea is to create an artificial column that responds the same way that biological ones do, i.e. built a virtual copy of an NCC and replicate it for real-world activity."
Coincidence? I don't think so. I stand by my original assertion.
H: Busted! Take your lumps & move on...Phil, you're friggin' amazing.
Erco, I think the paraphrasing you see is a good job as Phil pointed out, considering time constraints and other limitations (like being in and out of the hospital all week) and the informal nature of posts (i.e. this is not a published research paper, but simply for personal use only). However, as with any paraphrasing, no links are needed. In the future, I hope Phil can divert his resources to better use.
The new Brain 208 adds 208 Propeller Cogs to the Big Brain boosting it 123.81%. A key feature of 208 is the color coded wiring which makes assembly more efficient. After working with so many Propellers, the pins are memorized and tags are not necessarily needed. Also as a tip for early on wiring, a single tag is easily moved from chip to chip during wiring, like a key map.
______________________________________________
Introduction
As you know, there is an effort to increase the size of the Brain, giving it more processors and more Cogs, to satisfy the unending increasing hunger for more machine neurons. To help get started in this direction, another brain was created alongside the first, called the Twin Brain.
Former Twin Brain Status
The former robotic Twin Brain has now received an upgrade and is recycled - it's molded into a new Brain. There are now in effect two Brains. The entire concept of the Twin Brain was born as a neural matter injector and began with just two props on a single solderless breadboard. (see photos in previous posts)
Increasing Propeller Cogs
In an effort to raise the total number of Cogs in the Brain for a more maximized neural program handling, the board was expanded with additional Propellers. Propeller Cogs have increased from around 168 in the primary to 208 in the secondary. This increases the neural injection capability to 123.81%.
Functions of the 208 Brain
The new 208 Brain, which is no longer a twin, can serve as a Master Off-loader Machine as previously discussed. However, it is more likely that this much more massive appendage will be concatenated to the existing brain, thereby increasing neural connectivity. The connection brings the total Brain to over 376 processors.
208 Brain Testing
The 208 Brain is tested with LEDs on pin P15 located at the physical bottom of the chip. This facilitates wiring across the array in rows. The code and construction being tested is designed for both crystals and no crystals. This gives the option of high speed operations or low current consumption. Since the effort is currently to establish the greatest number of neural representations, complete Cog usage will be given a priority along with recycling the injector after its priming use and folding the neural matter.
Brain Matter Folding
As briefly discussed, Brain Matter Folding is extremely important and getting a 10x, 100x, 1000x, 10000x, and 1000000x folding result will be directly proportional to the number of elements of neural matter that the Brain can contain.
Neuron Definition
The neuron used in the machine brain is not the same as a neuron in the human brain, so references may or may not state "neural matter" to enable the flexibility required in a machine intelligence. Our neuron is currently a machine neuron and the neural matter is the substance and programming which surround and include it.
Brain Matter Folding Can you fold an electric brain?
Folding Intro
The human brain is capable of folding matter to achieve a high density of inborn material. This folding is actual physical folding. In the machine Brain, we will probably not (at this time) do any physical folding. The layout of the Exoskeleton (EXO) is already established in Flip Modes of operation and works well.
What Kind of Folding?
So what kind of folding can be applied to the numbers which comprise the neural matter? The numerical folding is compression. There are numerous applications for data compression with various algorithms. Which algorithm is best? To answer this question, a sample of the actual data and numerical material should be composed.
Types of Numerical Compression & Accuracy
Numerical data compression can include encryption, encoding, bandwidth manipulation, bit transposition, and care must be taken as not to incur loss, distortion, or inaccuracy. Just as an image which is JPG'ed too much and artifacts appear, it is the result of the artifacts that should be avoided, but if the compression can add up the number of black pixels and represent by a single number then recreate the image in its original perfect content, this would be an acceptable mode of compression. The interest is in lossless compression algorithms.
Decompression
Compression necessitates decompression which includes a time factor. Decompression can also be accomplished with hardware, although with the Brain, the focus is on software. A Propeller chip and its Cog could dedicate to this task of compression and decompression.
Running Length Algorithm
One example is running length algorithm. For example, 1955555 is 19(5)5 which is one nine and five fives. This compression code considers the data value and the length of the run.
Intelligence Compression Algorithm
Machine learning is capable of considering the entire history of data. This history can have either a sequence or an ordered arrangement that can be classified. Therefore it is possible to find the best route of compression based on sequential or historical data.
General Intelligence Algorithms
The proposal is for the creation of a general intelligence algorithm. Just as a human can memorize and implement a simple algorithm for deduction of the square root function, so can a machine. Just as the colors of the rainbow (Red, Orange, Yellow, Green, Blue, Indigo, and Violet) can appear from the compression jingle of Roy G. Biv, so can a machine's intelligence do the same.
The Progression Algorithm
Identification of progressions readily lend to compression.
The Matrices Algorithm
A combination of events, numbers, things, words, can fit the compression of a matrix.
... and the informal nature of posts (i.e. this is not a published research paper, but simply for personal use only). However, as with any paraphrasing, no links are needed.
That's nonsense! If it's just for personal use, why publish it here for the world to read? It doesn't matter how "informal" the venue. Without crediting the original author, you've represented someone else's words and ideas as your own. And that, sir, is known as plagiarism.
It makes total sense! Paraphrasing is absolutely not plagiarism, and does not require any reference to original authors. This is an open source hobby (personal use & enjoyment) development where anyone can join in worldwide. If anyone still thinks crediting is needed, be my guest, google every sentence and post sources. This dead dog is flogged enough. Time to put it to rest and move on.
That facebook site is blocked in China. I had no idea of its existence.
If you believe that paraphrasing without attribution is justified (and, frankly, some of the cribbed content goes beyond paraphrasing), why were you so keen to hide the fact that you had access to the selfsame article on sigularityhub.com that you claim was blocked on Facebook?
The New Brain 216 Third Brain Now Under Development
This is the beginning wiring of a new Brain 216 which is being built alongside Brain 208 and the first Big Brain 168. The combined raw brain power will make Brain 592 (216 + 208 + 168 = 592 processors). When Brain 216 is completed, another Twin Brain will be constructed. Brain 216 introduces a new high density breadboard that contains more Propeller chips per unit area. Slightly different wiring uses less breadboard holes. Connections for testing, leading from Brain 1 to Brain 2 to Brain 3 require only five main delivery wires.
The success of the second Propeller 208 Brain has spawned a new Third Brain. The new brain will have the same objective as the first and second Brains, i.e. to increase the total number of machine neurons and moving a step closer to machine intelligence. To do this, we need to continue the upward progression of more processors and more Cogs so these will be added on in the Third Brain.
The new wiring test bed on Brain 208 will be incorporated into the third brain. The new wiring test bed passed all tests thus far for crystal and non-crystal loading and burn in. Each test runs for a four hour burn in. New variations are added through programming. New wiring does not modify existing wiring significantly, i.e. it introduces some recycled pins and a new interface that is added on. The previous interface is not removed. The Hybrid is retained.
It may be a while until the new schematic is created. First, the full interface will be optimized as one phase of development. It's good the project is moving along rapidly. Next, the Third Brain will test parallel algorithms that were developed on the AM Algorithm Machine. This uses the HINT - Hybrid Interface.
Third Brain will use the color coding scheme for rapid assembly and line identifications, which match the first and second brains, and developed sub-brains.
Upon completion of Third Brain, it will be additive to Brain One and Brain Two. This will create one Big Brain. 208 Brain already takes up nearly half of the Lab bench. Third Brain may run in tandem, side by side on the other half of the bench.
The Parallel interface development will continue though lines for serial will be maintained. Eventually Guest Programmers will be able to gain time access to the primary Brain and run programming for intellectual development and to test various elements of machine intelligence.
The Brain program at that time may be operated like my remote robotic space telescope. Even though the telescope was remotely located, it could be robotic controlled by anyone in the world with the authority of access gain through internet. It is possible the Brain could be wired to internet through a Propeller chip for world wide access points in a similar manner as the remote robotic space telescope.
Remote programs could download into the Brain and execute by host programmers. More on this will be discussed in future posts. Input and ideas for remote program uploading to the Brain are welcomed.
Component status remains good - enough components exist to complete the Third Brain. Though the largest Brain version represents a sizable investment, a series of smaller brains are possible to fit any budget. These small brains will run the same programming, just on a smaller scale. More on these topics will follow.
Attribution is very important, whether for quoted or for paraphrased content -- or just for ideas you've come across. Not only does it give necessary credit to the original author where it's due, but it provides cues to your your readers that help them distinguish between your own ideas and those you've brought in from outside sources. If you develop a reputation for not giving due credit, then everything you post will carry the taint of suspicion: "Does he really understand what he's talking about, or is he just parroting someone else's ideas and buzzwords to bolster his own prestige?" None of us want that. Owning up to the error and making a visible effort to avoid it going forward will help to secure good "forum karma." Ignoring it or attempting to sweep it under the rug ... um ... not so much.
You've contributed a lot to this forum and, I'm sure, have many more good things to offer. For your own sake, please consider these matters carefully.
I've been skimming this thread on occasion, but not throughly reading it. So perhaps my next question has been answered.
But why do this in hardware?
The wiring challenge you are facing is called the tyranny of numbers. It becomes impractical past a certain point to hand wire any design with a large number of interconnected components. To avoid this problem, most neural networks I've seen are done in software using a honking big matrix. With a modern multi-core processor and multi-threaded programming you could simulate much large networks than you could build.
I've been skimming this thread on occasion, but not thoroughly reading it. So perhaps my next question has been answered. But why do this in hardware? The wiring challenge you are facing is called the tyranny of numbers. It becomes impractical past a certain point to hand wire any design with a large number of interconnected components. To avoid this problem, most neural networks I've seen are done in software using a honking big matrix. With a modern multi-core processor and multi-threaded programming you could simulate much large networks than you could build.
thanks Martin, good point.. here's my take on it..
Hardware Software Hybrid
Software Only
The point of doing similar projects through simulations on a computer through software only is of course a good idea and has much merit. Simulators were suggested before and have their place in the development of neural systems. These are "software-only" approaches to neural nets.
Hardware & Sofware
Other options include a mix of hardware and software. These nets use the best of two worlds, and may enable much higher speeds, accomplish more in unit time, and have hardware combined with the flexibility of software delivery systems.
Hardware
The third option uses hardware only to construct elements of the neural net, such as individual neurons. These systems can involve much hardware and can be less dynamic, requiring changes in wiring. On the other hand, static hardware models are useful for various forms of communications and the delivery and injection of neural matter.
Hybrid Combinations & The Brain
Any combination of these three approaches is possible. The current Brain project tries to include all three and uses a Hybrid arrangement.
Simulations were used to create programs on the PC that achieved and explored real time simulated neural nets. One of these plays a good game of Tic Tac Toe while another runs LIFE. (see previous posts)
Software is used to create neural matter and deliver it. Hardware is used to contain the neural matter, inject it, and distribute it.
Many of the neural connections are simply made in software and are delivered along a kind of synaptic pathway provided by the hardware in the forms of communication interfaces.
It's nothing too overly complicated, and using hardware where hardware is best suited, and software where software is best suited, keeps the project manageable.
The Other Factors
While simulating neural nets with software only is possible it negates the hardware experience. We can simulate Propeller chips on a pc but it is not the same as the actual hardware Propeller chip and misses the wiring experience and real world associations and learning experiences that accompany it.
There is an educational and learning factor to designing, hardware wiring and seeing it work. There is a hobby fun factor working with the hardware and software and enjoying the relaxing time of programming and hardware assembly.
There is the challenge of just doing it. The great exhilaration and extreme enjoyment, the adrenaline rush, the natural mind changing experiences, the fascination that exists in our hearts and souls... Why climb Mt. Everest, go to the Moon, discover new words, and build empty vessels to hold machine intelligence?
Some sources
[PDF]
NeuroLution: Integrated Hardware and Software for the Development ...
File Format: PDF/Adobe Acrobat - Quick View
by G Kock - Cited by 2 - Related articles
software and hardware tools for the development of neural network applications. However, a short overview of the state of the art allows to point out the ...
citeseerx.ist.psu.edu/viewdoc/download;jsessionid...?doi=10.1.1...
[PDF]
Neural Networks in Hardware: A Survey
man.ac.uk
File Format: PDF/Adobe Acrobat - Quick View
by Y Liao - Cited by 26 - Related articles
and run orders of magnitude faster than software simulations. Section 7 will present some real-world applications of neural network hardware. ...
Brain Control Panel
Progressing conceptual development
Like the cockpit of a Boeing 747, the proposed Brain command and control center can oversee many factors more than just uploading of programs.
This Brain 3C can not only monitor such things as temperature, stage, napping and dreaming, but look into higher ordered cognitive functions.
Like a kind of Thought Debug monitoring, the 3C will keep track of the Brain's well being. Access could include other areas:
vision
mobility
thought
nerve feelings
inputs
hearing
taste
touch
memory
indicators of communication
synaptic response
algorithm neural compression
decompression
Probably a Propeller chip would be dedicated to the BCP.
A utilities section would be useful with a speedometer to measure thought speed based on the type of process, the distance separating specific machine neurons and where they are located withing the neural grid, and the nature of multiple venues of machine intelligence.
It's possible a small version will undergo development, for testing the concept. There are many good examples for developing front end Propeller graphics with numbers, instrumentation, control panels etc.
New Brain Breadboard Design
Maximize breadboard space used
Introducing a new way to develop a new kind of breadboard, which is smaller, holds more Propeller chips, can fit more easily on the Lab Bench and is a really cool design! The workup achieves the smallest footprint while achieving the maximum in solderless breadboard Brain real estate.
This is the mother of all breadboards. You build up one, attach more, and have the highest density of Propeller chips per unit area possible in a solderless breadboard format.
This new breadboard design holds 216 cogs
and is expandable when connected to other
216 breadboards
Propellers are added to this design - going for
the maximum number of Propeller chips
without a Master Slave concept. This will
become a neural array assembly for testing and
concatenation to more neural arrays. It will
double as a massive parallel neural injector.
It will serve to test out new Brain algorithms.
The new breadboard for the latest Brain is designed with maximum space utilization in mind and simplicity in construction. You can buy lots of breadboards and assemble your own massively large board in the most efficient way.
How to make one..
Just score along the back side to remove the two power rails. Score out the notch area so the breadboard can be attached to one another. Connect eight breadboards together. Remove the backing and attach to the transparent plastic base. Add the two power rails at opposite side, and add the perpendicular board.
Add a power rail to each side
As you can see in the photo, to gain higher component density, some power bus strips are removed, some are redistributed, and a perpendicular board is added. A maximum number of breadboards connect along a single span. The two attached power bus rails are at opposing sides of the assembly. This unit can hold a total of 216 Cogs.
The boards connect to recycled clipboards of transparent plastic, a low cost material to work with and readily available at places like Wal-Mart. Peel off the backing to simply attach the boards.
Boards of course should be tested first. Some boards may have defective holes or empty holes. It depends on the quality of the board and the manufacturing company. For this high density design to work, all holes and connections must be in 100% order and available.
In some cases, a lot purchase of lower quality boards mean you must do the testing and throw away the boards that don't pass inspection. Higher quality boards are already tested and ready for use.
Multiple boards connect together for making larger Brains. Latest inventory shows a large supply of transparent plastic for making many more high density versions.
The wiring challenge you are facing is called the tyranny of numbers. It becomes impractical past a certain point to hand wire any design with a large number of interconnected components.
Humanoido, what has me scratching my head is the brain bus and distributed problem solving. Let's say you have 5 Propellers.
Prop 0 is taking distance readings (vision)
Prop 1 is controlling motor communication (mobility).
Prop 2 is listening for sounds/commands (hearing)
Prop 3 is controlling memory IO (memory)
Prop 4 is controlling touch (touch)
Do you have these concepts functioning in a real system? Can you provide example code that I can test drive?
Somewhere between then and now, Phil got me started on the Propeller chip. I started learning Spin and building larger machines that could do more. This led to another 20 machines of progressive size, that were torn down and recycled from one to the next, finally culminating in the largest UltraSpark series, going from 2 props on up to 40 props The data from the US40 led to the first 320 cog Brain experiment on Boe-Bot. This went so well (IMO), I immediately began hatching plans for a newer, leaner, more powerful brain.
Please post a link to the 320 cog Brain experiment on Boe-Bot code. How did you measure success on the 320 cog Brain experiment on Boe-Bot project?
The stock Boe-Bot code was loaded from the book and simply modified on-the-fly,
to slow down the servo movements so as not to topple the brain gantry.
The experiment achieved success and answered main questions.
1) Could a Boe-Bot carry a giant brain? (yes)
2) Could the US40 become the brain and operate in parallel? (yes)
3) Can we explore the concept of what it would take to make the Boe-Bot very smart? (yes)
4) Could it move by tether? (yes)
5) Can we investigate battery operations? (Yes)
The new Brain is a spinoff of this early experiment and has a Brain Stem - hardware designed to handle the motions directly. Test code is posted. The code will allow either sending or receiving a command to control motion along the BUS. A schematic is provided. It can operate on two platforms, a Propeller and a BS2. No other code for motion control is released at this time. However you can see the concept and start developing with it.
Humanoido, what has me scratching my head is the brain bus
and distributed problem solving. Let's say you have 5 Propellers.
Prop 0 is taking distance readings (vision)
Prop 1 is controlling motor communication (mobility).
Prop 2 is listening for sounds/commands (hearing)
Prop 3 is controlling memory IO (memory)
Prop 4 is controlling touch (touch)
Do you have these concepts functioning in a real system?
Can you provide example code that I can test drive?
The resources for those apps are on the OBEX. This Brain is going through many phases which represent various designs. The most recent design (see Brain Channeling) will allow the five props in the example to independently run in parallel at the same time in the brain.
Channels can collectively send data via channels to a specific gatherer propeller which may analyze the results and act upon it or merely display the results on LEDs, LCDs or TV. Is this a turn-key system? No.
But you are welcome to write the code. There is example code posted to send and receive data using the Brain Base and the Brain Spans for Propeller to Propeller communications. The BUS is also released as a schematic.
The Brain has additional interface lines for serial tx/rx - someone can develop some simple test software for this too. You may also experiment with WIDE BUS. My idea is to provide enough hardware connection for unique communication so programmers will be very happy.
I'm currently designing, shaping and tuning the neural matter injectors so that they may share a complete interface and recycle pins for communications. So this is ongoing development and it may take some considerable time before we have completed turn-key code.
I can steer you in a particular direction if you want to develop code for the project. Since we all have our own specific projects, I would suggest, that while doing your specific project - try to find a way to apply it to the Brain, with some changes.
In the past, I was able to successfully design code for one machine, and then adapt it to another, getting twice the mileage from one program. It's also possible to extract a pertinent subroutine from your larger code projects and build programs around it for additional apps.
The code used is found here:
Robotics with the Boe-Bot Text v3.0 (.pdf)
The stock Boe-Bot code was loaded from the book and simply modified on-the-fly,
to slow down the servo movements so as not to topple the brain gantry.
Can you see how this statement is confusing?
The data from the US40 led to the first 320 cog Brain experiment on Boe-Bot. This went so well (IMO), I immediately began hatching plans for a newer, leaner, more powerful brain.
The Brain has additional interface lines for serial tx/rx - someone can develop some simple test software for this too. You may also experiment with WIDE BUS. My idea is to provide enough hardware connection for unique communication so programmers will be very happy.
The experiment has nothing to do with distributing Boe-Bot functions across the US40 Prop array. But you state that the experimental was a success so a new Brain was developed.
The resources for those apps are on the OBEX. This Brain is going through many phases which represent various designs. The most recent design (see Brain Channeling) will allow the five props in the example to independently run in parallel at the same time in the brain.
Where is the Brain Channeling in the OBEX?
But you are welcome to write the code. There is example code posted to send and receive data using the Brain Base and the Brain Spans for Propeller to Propeller communications. The BUS is also released as a schematic.
I beleive you're are referencing the 1/2 duplex and full duplex code? This architecture will work fine in Prop to Prop or Prop to STAMP communication but not so well in a large parallel network. For contribution, I wrote a 1Mpbs 1/2 duplex code that frames a command packet and shoots it off to a mutli-drop serial bus. It's in the OBEX in the AX-12 projects. Still, that would not work well in a parallel system. It's a master slave setup.
I'm currently designing, shaping and tuning the neural matter injectors so that they may share a complete interface and recycle pins for communications. So this is ongoing development and it may take some considerable time before we have completed turn-key code.
What is a neural matter injector? Do you have a code example? How do you "recycle pins for communications"? Does this mean multi-purpose IO?
I can steer you in a particular direction if you want to develop code for the project. Since we all have our own specific projects, I would suggest, that while doing your specific project - try to find a way to apply it to the Brain, with some changes.
In the past, I was able to successfully design code for one machine, and then adapt it to another, getting twice the mileage from one program. It's also possible to extract a pertinent subroutine from your larger code projects and build programs around it for additional apps.
All my Prop projects have a similar architecture where data is passed through the HUB. Nothing new just using the Propeller as intended. I have no idea how to write code for the Brain as the interface is not defined. To contribute to the project, I would expect clear specifications that outline how code must behave in the brain. Things like basic object communication within a Prop and with external Props. Without specifications I'd be wiring the Brain from scratch.
Sorry about that. I should write more information. However, it's very clear to me. What part seems confusing?
>The experiment has nothing to do with distributing Boe-Bot functions across the US40 Prop array.
This is correct. The US40 was doing its own brain thinking. I think there was an experimental PING on it at that time for vision, but it was later removed.
In the post above, the objectives of the Boe-Bot experiment are listed. They are copied here again.
The experiment achieved success and answered main questions.
1) Could a Boe-Bot carry a giant brain? (yes)
2) Could the US40 become the brain and operate in parallel? (yes)
3) Can we explore the concept of what it would take to make the Boe-Bot very smart? (yes)
4) Could it move by tether? (yes)
5) Can we investigate battery operations? (Yes)
>But you state that the experimental was a success so a new Brain was developed.
Yes, the points of the experiment listed above were all successful. Without investigating those ideas, it would be difficult to progress to the next generation brain.
Comments
Using this Brain Twin and the largest NMI, the
Brain will be transformed into the largest
machine.
This first NMI was recycled into a much
larger unit. It could attach to the Brain and
the Brain Twin to create the largest Humanoido
Machine and Brain to date,
If (and when) the primary Big Brain is connected to the Twin Brain and connected to the neural matter injector that was originally completed, the size of the Big Brain will exceed the size of the largest Humanoido project built to date, which is the UltraSpark 40, a 320 Propeller Processor machine.
In terms of power, which should be defined first - is that quest to engage (among other things) the maximum amount of neural containment, also classified further as neural substance, neural matter, and neurons.
Note: in our definitions of neurons, we will define a machine neuron and not that which comprises a human neuron. Machine neurons can be manufactured to dilate once injected, thus lowering the numerical requirement for a specific brain containment.
The objective of the Big Brain is not to replicate the details of the Human Brain (unless absolutely necessary for some reason), but rather only use the brain as an outline and reference.
Dilation is an algorithm that will be explored when the point of neural propagation is reached.
In the past, a rotational matrix was used for conditioning containment fields, and so if any budding mathematician programmers want to pave the path for neural matter dilation in a Propeller chips propagation of neural matter, go ahead. A matrix equation could be a future topic.
Neural Matter Injectors are multi-purposed
All NMI's - Neural Matter Injectors can have multiple functions
I am now convinced that machine life can be established with the Brain. In studying the methods and developing methods to lead to this objective, it becomes very important to define exactly what constitutes life, and what comprises its various components.
We are probably not interested in conception and propagation of little brains as procreated by the big brain, though reproduction may be in someone's definition of life. Yet, there are many people in life who, for one reason or another, cannot have children. These people are obviously "alive."
Also as pointed our earlier, if you cannot turn the door knob to open the door, your level of intelligence can still be off the high end scale of IQ.
Yet, it is not often that we have opportunity to meet genius, and when my Physics professor ribbed me hard during the Astrophysics conference in Lincoln NE, I looked up only to discover that one of the world's greatest Astrophysicists of all time, MIT's Professor Morrison, was sitting right next to me!
During that day, he taught things about the Universe that would never be forgotten - but I learned something else, that in his quest of genius level understanding, his body devoted all its physical resources to his brain. This indicates that devoting lots of resources to the machine brain will be important for the highest level of intelligence.
So in the light of the doorknob, it seems relatively unimportant when it comes to a level understanding of gravitational evolution and converse devolution. Yet, based again on a definition, the turning of the doorknob may be a priority.
On to the Key. What contributes to making a machine alive is its given purpose, programming, routines, heuristic algorithms, subroutines, level of programming, architecture and application. In a machine, we can bring forth different levels of life, styles of life, let the machine have leeway to make decisions and learn, set the rules, be inquisitive, and govern the applications and set the domains.
It's about the life of anything that can be understood and defined, and then transformed into the machine. We are the master, but one school of thought is we teach the machine and let it learn beyond its programming. Again some definitions are needed. What is "beyond its programming?" This may be as simple as extending a given database, asking questions that were not specifically programmed, and having some originality or creativity when painting.
Again, as with painting, we are dealing with a program, is it water based, oil based, charcoal, pencil, color, B&W, conceptual materials, etc. and what is the style of painting, i.e. is it like a Picasso, Rembrandt, Monet, da Vinci, Van Gogh, or like a machine?
I suggest we look at different levels of intelligence base on varying applications. Again, we need to define applications in some highly specific ways.
Mike G and Humanoido make good points about the various challenges for a machine to open a door it (not ready to call them he/she yet) hasn't encountered before.
I believe a lot of thought has gone in to teaching robots to climb stairs and open doors. It seems it's easier to teach a robot to fly than to teach one to travel up a couple of flights of stairs and turn a door knob.
As you may know, while not a primary objective to the "Fill the Brain" project, there is some use in reverse-engineering the human brain, or animal brains - not as a whole but as a fractional part.
The human brain is made up of around 100 billion neurons. These neurons connect and communicate with each other through a massive network of around 100 trillion synapses. This indicates the great complexity required to make a human artificial brain.
There's a trick that the human body uses. Genes don’t map out every single detail - they give a more general instruction and repeat a few million times (i.e. Fractal Instructions).
One such repeating structural unit is called a Neocortical Column (NCC) comprised of a group of about 10,000 neurons in the Cerebral Cortex, organized in a relatively consistent way across a brain. Grey matter contains millions of these columns.
Rather than trying to create a model of the whole brain at once, one idea is to create an artificial column that responds the same way that biological ones do, i.e. built a virtual copy of an NCC and replicate it for real-world activity.
Thus, the current brain crop can handle two Neo Cortical Columns at 20,000. The 1st NCC can be cloned, just as we clone the neurons. Just what we can do with two of these columns remains to be seen.
However, what we can learn and borrow from the NCC is the general repeating function. If similar compacting ratios can be achieved, i.e. one million to one, it could have serious impact and implications on the machine brain.
"Fractal Instructions?" You're just parroting buzzwords. When you quote or paraphrase someone else's work, you -- at the very least -- need to give credit, viz:
-Phil
Fractal Dimension in Human Cerebellum Measured by Magnetic Resonance Imaging
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1303704/
Fractal Geometry, is a good introduction (chapter 7)
http://www.google.com/url?sa=t&source=web&cd=8&ved=0CFQQFjAH&url=http%3A%2F%2Fwww.quantumk.co.uk%2Fquantumk%2FChapter%25207%2520-%2520Fractal%2520geometry.pdf&rct=j&q=fractal%20instructions%20human%20brain&ei=t6WWTf7xOIaMvQPVltDjCw&usg=AFQjCNE6xDutfevK04lo3XJ8OOPhdLBRzA&cad=rja
Fractals in the nervous system: conceptual implications for theoretical neuroscience
http://www.frontiersin.org/fractal_physiology/10.3389/fphys.2010.00015/full
Phil, can you offer a simple demo fractal folding program based on a simple neuron?
-Phil
Humanoido: "Genes don’t map out every single detail - they give a more general instruction and repeat a few million times (i.e. Fractal Instructions)."
BBP: "One such structure is called a neocortical column (NCC): a group of about 10,000 neurons in the cerebral cortex that are organized in a relatively consistent way across the mammalian brain. Millions of these columns compose the whole of your grey matter."
Humanoido: "One such repeating structural unit is called a Neocortical Column (NCC) comprised of a group of about 10,000 neurons in the Cerebral Cortex, organized in a relatively consistent way across a brain. Grey matter contains millions of these columns."
BBP: "Rather than trying to create a model of the whole brain at once, the Blue Brain Project is attempting to accurately model a single NCC in a rat’s brain. If they can create an artificial column that responds the same way that biological ones do to electrical impulses, they’ll be on the right track to building a good model."
Humanoido: "Rather than trying to create a model of the whole brain at once, one idea is to create an artificial column that responds the same way that biological ones do, i.e. built a virtual copy of an NCC and replicate it for real-world activity."
Coincidence? I don't think so. I stand by my original assertion.
-Phil
Phil, you're friggin' amazing.
Another robot Brain is born
The new Brain 208 adds 208 Propeller Cogs to the Big Brain boosting it 123.81%. A key feature of 208 is the color coded wiring which makes assembly more efficient. After working with so many Propellers, the pins are memorized and tags are not necessarily needed. Also as a tip for early on wiring, a single tag is easily moved from chip to chip during wiring, like a key map.
______________________________________________
Introduction
As you know, there is an effort to increase the size of the Brain, giving it more processors and more Cogs, to satisfy the unending increasing hunger for more machine neurons. To help get started in this direction, another brain was created alongside the first, called the Twin Brain.
Former Twin Brain Status
The former robotic Twin Brain has now received an upgrade and is recycled - it's molded into a new Brain. There are now in effect two Brains. The entire concept of the Twin Brain was born as a neural matter injector and began with just two props on a single solderless breadboard. (see photos in previous posts)
Increasing Propeller Cogs
In an effort to raise the total number of Cogs in the Brain for a more maximized neural program handling, the board was expanded with additional Propellers. Propeller Cogs have increased from around 168 in the primary to 208 in the secondary. This increases the neural injection capability to 123.81%.
Functions of the 208 Brain
The new 208 Brain, which is no longer a twin, can serve as a Master Off-loader Machine as previously discussed. However, it is more likely that this much more massive appendage will be concatenated to the existing brain, thereby increasing neural connectivity. The connection brings the total Brain to over 376 processors.
208 Brain Testing
The 208 Brain is tested with LEDs on pin P15 located at the physical bottom of the chip. This facilitates wiring across the array in rows. The code and construction being tested is designed for both crystals and no crystals. This gives the option of high speed operations or low current consumption. Since the effort is currently to establish the greatest number of neural representations, complete Cog usage will be given a priority along with recycling the injector after its priming use and folding the neural matter.
Brain Matter Folding
As briefly discussed, Brain Matter Folding is extremely important and getting a 10x, 100x, 1000x, 10000x, and 1000000x folding result will be directly proportional to the number of elements of neural matter that the Brain can contain.
Neuron Definition
The neuron used in the machine brain is not the same as a neuron in the human brain, so references may or may not state "neural matter" to enable the flexibility required in a machine intelligence. Our neuron is currently a machine neuron and the neural matter is the substance and programming which surround and include it.
Can you fold an electric brain?
Folding Intro
The human brain is capable of folding matter to achieve a high density of inborn material. This folding is actual physical folding. In the machine Brain, we will probably not (at this time) do any physical folding. The layout of the Exoskeleton (EXO) is already established in Flip Modes of operation and works well.
What Kind of Folding?
So what kind of folding can be applied to the numbers which comprise the neural matter? The numerical folding is compression. There are numerous applications for data compression with various algorithms. Which algorithm is best? To answer this question, a sample of the actual data and numerical material should be composed.
Types of Numerical Compression & Accuracy
Numerical data compression can include encryption, encoding, bandwidth manipulation, bit transposition, and care must be taken as not to incur loss, distortion, or inaccuracy. Just as an image which is JPG'ed too much and artifacts appear, it is the result of the artifacts that should be avoided, but if the compression can add up the number of black pixels and represent by a single number then recreate the image in its original perfect content, this would be an acceptable mode of compression. The interest is in lossless compression algorithms.
Decompression
Compression necessitates decompression which includes a time factor. Decompression can also be accomplished with hardware, although with the Brain, the focus is on software. A Propeller chip and its Cog could dedicate to this task of compression and decompression.
Running Length Algorithm
One example is running length algorithm. For example, 1955555 is 19(5)5 which is one nine and five fives. This compression code considers the data value and the length of the run.
Intelligence Compression Algorithm
Machine learning is capable of considering the entire history of data. This history can have either a sequence or an ordered arrangement that can be classified. Therefore it is possible to find the best route of compression based on sequential or historical data.
General Intelligence Algorithms
The proposal is for the creation of a general intelligence algorithm. Just as a human can memorize and implement a simple algorithm for deduction of the square root function, so can a machine. Just as the colors of the rainbow (Red, Orange, Yellow, Green, Blue, Indigo, and Violet) can appear from the compression jingle of Roy G. Biv, so can a machine's intelligence do the same.
The Progression Algorithm
Identification of progressions readily lend to compression.
The Matrices Algorithm
A combination of events, numbers, things, words, can fit the compression of a matrix.
-Phil
-Phil
Third Brain Now Under Development
This is the beginning wiring of a new Brain 216 which is being built alongside Brain 208 and the first Big Brain 168. The combined raw brain power will make Brain 592 (216 + 208 + 168 = 592 processors). When Brain 216 is completed, another Twin Brain will be constructed. Brain 216 introduces a new high density breadboard that contains more Propeller chips per unit area. Slightly different wiring uses less breadboard holes. Connections for testing, leading from Brain 1 to Brain 2 to Brain 3 require only five main delivery wires.
___________________________________________________
The success of the second Propeller 208 Brain has spawned a new Third Brain. The new brain will have the same objective as the first and second Brains, i.e. to increase the total number of machine neurons and moving a step closer to machine intelligence. To do this, we need to continue the upward progression of more processors and more Cogs so these will be added on in the Third Brain.
The new wiring test bed on Brain 208 will be incorporated into the third brain. The new wiring test bed passed all tests thus far for crystal and non-crystal loading and burn in. Each test runs for a four hour burn in. New variations are added through programming. New wiring does not modify existing wiring significantly, i.e. it introduces some recycled pins and a new interface that is added on. The previous interface is not removed. The Hybrid is retained.
It may be a while until the new schematic is created. First, the full interface will be optimized as one phase of development. It's good the project is moving along rapidly. Next, the Third Brain will test parallel algorithms that were developed on the AM Algorithm Machine. This uses the HINT - Hybrid Interface.
http://forums.parallax.com/showthread.php?124433-Tiny-Tester-for-Developing-Parallel-Algorithms
Third Brain will use the color coding scheme for rapid assembly and line identifications, which match the first and second brains, and developed sub-brains.
Upon completion of Third Brain, it will be additive to Brain One and Brain Two. This will create one Big Brain. 208 Brain already takes up nearly half of the Lab bench. Third Brain may run in tandem, side by side on the other half of the bench.
The Parallel interface development will continue though lines for serial will be maintained. Eventually Guest Programmers will be able to gain time access to the primary Brain and run programming for intellectual development and to test various elements of machine intelligence.
The Brain program at that time may be operated like my remote robotic space telescope. Even though the telescope was remotely located, it could be robotic controlled by anyone in the world with the authority of access gain through internet. It is possible the Brain could be wired to internet through a Propeller chip for world wide access points in a similar manner as the remote robotic space telescope.
Remote programs could download into the Brain and execute by host programmers. More on this will be discussed in future posts. Input and ideas for remote program uploading to the Brain are welcomed.
Component status remains good - enough components exist to complete the Third Brain. Though the largest Brain version represents a sizable investment, a series of smaller brains are possible to fit any budget. These small brains will run the same programming, just on a smaller scale. More on these topics will follow.
Here is a site with some helpful rules for avoiding plagiarism:
Attribution is very important, whether for quoted or for paraphrased content -- or just for ideas you've come across. Not only does it give necessary credit to the original author where it's due, but it provides cues to your your readers that help them distinguish between your own ideas and those you've brought in from outside sources. If you develop a reputation for not giving due credit, then everything you post will carry the taint of suspicion: "Does he really understand what he's talking about, or is he just parroting someone else's ideas and buzzwords to bolster his own prestige?" None of us want that. Owning up to the error and making a visible effort to avoid it going forward will help to secure good "forum karma." Ignoring it or attempting to sweep it under the rug ... um ... not so much.
You've contributed a lot to this forum and, I'm sure, have many more good things to offer. For your own sake, please consider these matters carefully.
Thanks,
-Phil
But why do this in hardware?
The wiring challenge you are facing is called the tyranny of numbers. It becomes impractical past a certain point to hand wire any design with a large number of interconnected components. To avoid this problem, most neural networks I've seen are done in software using a honking big matrix. With a modern multi-core processor and multi-threaded programming you could simulate much large networks than you could build.
thanks Martin, good point.. here's my take on it..
Hardware Software Hybrid
Software Only
The point of doing similar projects through simulations on a computer through software only is of course a good idea and has much merit. Simulators were suggested before and have their place in the development of neural systems. These are "software-only" approaches to neural nets.
Hardware & Sofware
Other options include a mix of hardware and software. These nets use the best of two worlds, and may enable much higher speeds, accomplish more in unit time, and have hardware combined with the flexibility of software delivery systems.
Hardware
The third option uses hardware only to construct elements of the neural net, such as individual neurons. These systems can involve much hardware and can be less dynamic, requiring changes in wiring. On the other hand, static hardware models are useful for various forms of communications and the delivery and injection of neural matter.
Hybrid Combinations & The Brain
Any combination of these three approaches is possible. The current Brain project tries to include all three and uses a Hybrid arrangement.
Simulations were used to create programs on the PC that achieved and explored real time simulated neural nets. One of these plays a good game of Tic Tac Toe while another runs LIFE. (see previous posts)
Software is used to create neural matter and deliver it. Hardware is used to contain the neural matter, inject it, and distribute it.
Many of the neural connections are simply made in software and are delivered along a kind of synaptic pathway provided by the hardware in the forms of communication interfaces.
It's nothing too overly complicated, and using hardware where hardware is best suited, and software where software is best suited, keeps the project manageable.
The Other Factors
While simulating neural nets with software only is possible it negates the hardware experience. We can simulate Propeller chips on a pc but it is not the same as the actual hardware Propeller chip and misses the wiring experience and real world associations and learning experiences that accompany it.
There is an educational and learning factor to designing, hardware wiring and seeing it work. There is a hobby fun factor working with the hardware and software and enjoying the relaxing time of programming and hardware assembly.
There is the challenge of just doing it. The great exhilaration and extreme enjoyment, the adrenaline rush, the natural mind changing experiences, the fascination that exists in our hearts and souls... Why climb Mt. Everest, go to the Moon, discover new words, and build empty vessels to hold machine intelligence?
Some sources
[PDF]
NeuroLution: Integrated Hardware and Software for the Development ...
File Format: PDF/Adobe Acrobat - Quick View
by G Kock - Cited by 2 - Related articles
software and hardware tools for the development of neural network applications. However, a short overview of the state of the art allows to point out the ...
citeseerx.ist.psu.edu/viewdoc/download;jsessionid...?doi=10.1.1...
[PDF]
Neural Networks in Hardware: A Survey
man.ac.uk
File Format: PDF/Adobe Acrobat - Quick View
by Y Liao - Cited by 26 - Related articles
and run orders of magnitude faster than software simulations. Section 7 will present some real-world applications of neural network hardware. ...
Progressing conceptual development
Like the cockpit of a Boeing 747, the proposed Brain command and control center can oversee many factors more than just uploading of programs.
This Brain 3C can not only monitor such things as temperature, stage, napping and dreaming, but look into higher ordered cognitive functions.
Like a kind of Thought Debug monitoring, the 3C will keep track of the Brain's well being. Access could include other areas:
- vision
- mobility
- thought
- nerve feelings
- inputs
- hearing
- taste
- touch
- memory
- indicators of communication
- synaptic response
- algorithm neural compression
- decompression
Probably a Propeller chip would be dedicated to the BCP.A utilities section would be useful with a speedometer to measure thought speed based on the type of process, the distance separating specific machine neurons and where they are located withing the neural grid, and the nature of multiple venues of machine intelligence.
It's possible a small version will undergo development, for testing the concept. There are many good examples for developing front end Propeller graphics with numbers, instrumentation, control panels etc.
Maximize breadboard space used
Introducing a new way to develop a new kind of breadboard, which is smaller, holds more Propeller chips, can fit more easily on the Lab Bench and is a really cool design! The workup achieves the smallest footprint while achieving the maximum in solderless breadboard Brain real estate.
This is the mother of all breadboards. You build up one, attach more, and have the highest density of Propeller chips per unit area possible in a solderless breadboard format.
This new breadboard design holds 216 cogs
and is expandable when connected to other
216 breadboards
Propellers are added to this design - going for
the maximum number of Propeller chips
without a Master Slave concept. This will
become a neural array assembly for testing and
concatenation to more neural arrays. It will
double as a massive parallel neural injector.
It will serve to test out new Brain algorithms.
The new breadboard for the latest Brain is designed with maximum space utilization in mind and simplicity in construction. You can buy lots of breadboards and assemble your own massively large board in the most efficient way.
How to make one..
Just score along the back side to remove the two power rails. Score out the notch area so the breadboard can be attached to one another. Connect eight breadboards together. Remove the backing and attach to the transparent plastic base. Add the two power rails at opposite side, and add the perpendicular board.
Add a power rail to each side
As you can see in the photo, to gain higher component density, some power bus strips are removed, some are redistributed, and a perpendicular board is added. A maximum number of breadboards connect along a single span. The two attached power bus rails are at opposing sides of the assembly. This unit can hold a total of 216 Cogs.
The boards connect to recycled clipboards of transparent plastic, a low cost material to work with and readily available at places like Wal-Mart. Peel off the backing to simply attach the boards.
Boards of course should be tested first. Some boards may have defective holes or empty holes. It depends on the quality of the board and the manufacturing company. For this high density design to work, all holes and connections must be in 100% order and available.
In some cases, a lot purchase of lower quality boards mean you must do the testing and throw away the boards that don't pass inspection. Higher quality boards are already tested and ready for use.
Multiple boards connect together for making larger Brains. Latest inventory shows a large supply of transparent plastic for making many more high density versions.
The current version offers:
Attach a side perpendicular board
Prop 0 is taking distance readings (vision)
Prop 1 is controlling motor communication (mobility).
Prop 2 is listening for sounds/commands (hearing)
Prop 3 is controlling memory IO (memory)
Prop 4 is controlling touch (touch)
Do you have these concepts functioning in a real system? Can you provide example code that I can test drive?
Please post a link to the 320 cog Brain experiment on Boe-Bot code. How did you measure success on the 320 cog Brain experiment on Boe-Bot project?
More on the Smartest Boe-Bot Project
1st Boe-Bot Brain appendage
Here's two links to the project.
http://forums.parallax.com/showthread.php?123909-Smartest-BoeBot
http://forums.parallax.com/showthread.php?123828-40-Props-in-a-Skyscraper
The code used is found here:
Robotics with the Boe-Bot Text v3.0 (.pdf)
The stock Boe-Bot code was loaded from the book and simply modified on-the-fly,
to slow down the servo movements so as not to topple the brain gantry.
The experiment achieved success and answered main questions.
1) Could a Boe-Bot carry a giant brain? (yes)
2) Could the US40 become the brain and operate in parallel? (yes)
3) Can we explore the concept of what it would take to make the Boe-Bot very smart? (yes)
4) Could it move by tether? (yes)
5) Can we investigate battery operations? (Yes)
The new Brain is a spinoff of this early experiment and has a Brain Stem - hardware designed to handle the motions directly. Test code is posted. The code will allow either sending or receiving a command to control motion along the BUS. A schematic is provided. It can operate on two platforms, a Propeller and a BS2. No other code for motion control is released at this time. However you can see the concept and start developing with it.
The resources for those apps are on the OBEX. This Brain is going through many phases which represent various designs. The most recent design (see Brain Channeling) will allow the five props in the example to independently run in parallel at the same time in the brain.
Channels can collectively send data via channels to a specific gatherer propeller which may analyze the results and act upon it or merely display the results on LEDs, LCDs or TV. Is this a turn-key system? No.
But you are welcome to write the code. There is example code posted to send and receive data using the Brain Base and the Brain Spans for Propeller to Propeller communications. The BUS is also released as a schematic.
The Brain has additional interface lines for serial tx/rx - someone can develop some simple test software for this too. You may also experiment with WIDE BUS. My idea is to provide enough hardware connection for unique communication so programmers will be very happy.
I'm currently designing, shaping and tuning the neural matter injectors so that they may share a complete interface and recycle pins for communications. So this is ongoing development and it may take some considerable time before we have completed turn-key code.
I can steer you in a particular direction if you want to develop code for the project. Since we all have our own specific projects, I would suggest, that while doing your specific project - try to find a way to apply it to the Brain, with some changes.
In the past, I was able to successfully design code for one machine, and then adapt it to another, getting twice the mileage from one program. It's also possible to extract a pertinent subroutine from your larger code projects and build programs around it for additional apps.
Can you see how this statement is confusing? The experiment has nothing to do with distributing Boe-Bot functions across the US40 Prop array. But you state that the experimental was a success so a new Brain was developed.
Where is the Brain Channeling in the OBEX?
I beleive you're are referencing the 1/2 duplex and full duplex code? This architecture will work fine in Prop to Prop or Prop to STAMP communication but not so well in a large parallel network. For contribution, I wrote a 1Mpbs 1/2 duplex code that frames a command packet and shoots it off to a mutli-drop serial bus. It's in the OBEX in the AX-12 projects. Still, that would not work well in a parallel system. It's a master slave setup.
What is a neural matter injector? Do you have a code example? How do you "recycle pins for communications"? Does this mean multi-purpose IO?
All my Prop projects have a similar architecture where data is passed through the HUB. Nothing new just using the Propeller as intended. I have no idea how to write code for the Brain as the interface is not defined. To contribute to the project, I would expect clear specifications that outline how code must behave in the brain. Things like basic object communication within a Prop and with external Props. Without specifications I'd be wiring the Brain from scratch.
>Can you see how this statement is confusing?
Sorry about that. I should write more information. However, it's very clear to me. What part seems confusing?
>The experiment has nothing to do with distributing Boe-Bot functions across the US40 Prop array.
This is correct. The US40 was doing its own brain thinking. I think there was an experimental PING on it at that time for vision, but it was later removed.
In the post above, the objectives of the Boe-Bot experiment are listed. They are copied here again.
The experiment achieved success and answered main questions.
1) Could a Boe-Bot carry a giant brain? (yes)
2) Could the US40 become the brain and operate in parallel? (yes)
3) Can we explore the concept of what it would take to make the Boe-Bot very smart? (yes)
4) Could it move by tether? (yes)
5) Can we investigate battery operations? (Yes)
>But you state that the experimental was a success so a new Brain was developed.
Yes, the points of the experiment listed above were all successful. Without investigating those ideas, it would be difficult to progress to the next generation brain.