Big Brain is now functional. ... Big Brain can ... learn, evolve, demonstrate Life, has neurons, ... In tests, software loads over 100,000 exampling neurons and can inject a variety of sample neurons.
This is great news, indeed! I guess this means that you now have some source code for those "neurons" that you can share with us, right?
This is great news, indeed! I guess this means that you now have some source code for those "neurons" that you can share with us, right?
-Phil
There was a neuron object posted some time ago. Seemed like it was a neuron per chip and used the float32 object.
It is possible to make a "node set" with a propeller containing say 28 neurons using scaled integer math. Each "Neural COG" would have to run multiple calculators and signal a storage COG for saving states perhaps in EEPROM. Each neuron could have 4 tokenized "streams of conciousness." How to merge all of that into sentient being is another question.
A brain should pass the Turing test, otherwise it is just a collage of objects.
Apparently there are a lot of collages walking around, chewing gum, making money, driving cars, sipping wine, running for office.
Honestly, I thought the Turning Test was shrugged off years ago. It requires a judge of the test to be fooled - or not. But who is that judge? And who judges the judge to see if the judge is competent to judge? And who judges the judges of the judge, etc, etc, ad infinitum?
I have no idea what the heck Humanoido is doing. But, in any case, it looks like fun.
There was a neuron object posted some time ago. Seemed like it was a neuron per chip and used the float32 object. It is possible to make a "node set" with a propeller containing say 28 neurons using scaled integer math. Each "Neural COG" would have to run multiple calculators and signal a storage COG for saving states perhaps in EEPROM. Each neuron could have 4 tokenized "streams of conciousness." How to merge all of that into sentient being is another question.
...there might be a few who wish to explore this concept.
The idea that I have posted is basically a huge 'ring' FIFO shift register with several taps that can either be used as an input or output. When used as an input, external events are allowed to be 'inserted' into the ring creating an 'echo' of the actual events that propagate around the ring. When used as an output, the idea is that you just read the status of the data taps.
To create a flux within the net each bit that is read is stored as the inverse of what was actually read... i.e a '0' is stored as a '1' and a '1' is stored as a '0'. A 'quiet' undisturbed net should have equal number of bits high versus equal number of bits low at any given moment in time. Any external influence on the net should bias the net away from 50/50 and it should remain biased within the ring even after the stimulus has been removed. In this sense the event has been recorded to the 'neural net' cell and new events can use predictive comparisons against previously stored events.
"Beau how do you propose something useful as a standalone device would work with your suggestion?"
That is almost like handing someone in the 1800's an airplane wing and asking what you could possibly use it for.
However, something of great interest to me, in answer to your question would be to setup a 'neural-net-engine'... think if each bit progressed internally on each prop building up a histogram for each pin. outside influences (sensors, even triggers) could affect some bits. Over time many 'F' generations back-n-forth between the props you would have a pattern based on the histogram of an event or series of events. This could be broken down into individual frames, or represent large spans of time. A particular setup like this between two propellers wouldn't need HUGE amounts of memory. This just provides an engine, that perhaps other (read one or several) propellers could tie into the 'neural-net-engine' and contribute.
Think of a 'Chair', a simple chair. You had to be told that it was a chair right, but now subconsciously, you know when you see a chair, only because you have identified it with some kind of reference 'data' or histogram. Now imagine several chairs, where each time the histogram builds upon the previous data, and only the common mode information that is pertinent remains.
Beau Schwabe
Parallax Semiconductor IC Layout Engineer
"Big Brain is now functional. ... Big Brain can ... learn, evolve, demonstrate Life, has neurons, ... In tests, software loads over 100,000 exampling neurons and can inject a variety of sample neurons."
Yet, when pressed about actual running software, he quotes an abstract concept by Beau that may never have been fleshed out. I'm all for fanciful art installations, and it's fun to construct "castles in the sky"; but when claims are made of actual performance, I think we deserve to see details that can be replicated.
Where is the Propeller code for these "neurons", Humanoido? "Enquiring Minds Want to Know." TM
I want to see some big brain rocket experiments...
Ttailspin, as you can see in the photo, Phil and his volunteer Propeller team are working around the clock on this homebuilt rocket spacecraft to launch the brain. As Phil said, "see that big black hole? That's where you stick your brain."
From left to right: Phil attaching his one of his many stars. Jazzed: a little hammered. ElectricAye crow bar adjusting the nose cone and inscribing Calculus & differential flight equations. Mike G in the suit with pre flight jitters holding the remote. Fido looks on.
I think this quote puts much of what is needed from the big brain into perspective...
Think of a 'Chair', a simple chair. You had to be told that it was a chair right, but now subconsciously, you know when you see a chair, only because you have identified it with some kind of reference 'data' or histogram. Now imagine several chairs, where each time the histogram builds upon the previous data, and only the common mode information that is pertinent remains.
Beau Schwabe Parallax Semiconductor IC Layout Engineer
"Big Brain is now functional. ... Big Brain can ... learn, evolve, demonstrate Life, has neurons, ... In tests, software loads over 100,000 exampling neurons and can inject a variety of sample neurons."
Yet, when pressed about actual running software, he quotes an abstract concept by Beau that may never have been fleshed out. I'm all for fanciful art installations, and it's fun to construct "castles in the sky"; but when claims are made of actual performance, I think we deserve to see details that can be replicated.
Where is the Propeller code for these "neurons", Humanoido? "Enquiring Minds Want to Know." TM
The Big Brain is being prepped as a robot. It already has a Brain Stem (see index) dedicated to motion control and will be outfitted with servos, mechanics and various limbs.
It's a constituent part of a robot, a robot brain. A refined condensed version will be used with humanoid robots. Smaller versions of the Big Brain are being prepared for smaller robots. This is a large ongoing project with puzzle pieces that fit together over time but won't be fleshed out until completion.
The taxonomic hierarchy of the type used by Collins and Quillian in their model of the organization of knowledge in memory appears very similar to the programmed behaviors of Lerner, which is learning code already completed and running on the Propeller.
The taxonomic hierarchy of the type used by Collins and Quillian in their model of the organization of knowledge in memory appears very similar to the programmed behaviors of Lerner, which is learning code already completed and running on the Propeller.
Phil Pilgrim, thanks I can see you're very interested and feel like an impatient child waiting for Santa's gifts on Christmas eve... Big Brain says, "Code is expected to post upon project completion. Have patience."
Originally Posted by Humanoido
The taxonomic hierarchy of the type used by Collins and Quillian in their model of the organization of knowledge in memory appears very similar to the programmed behaviors of Lerner, which is learning code already completed and running on the Propeller.
Download the code in the Big Brain source found below: The Big Brain Thread, Page 27, post number 523, Machine Learning with Lerner - Dave Hein fills a small brain! Dave has since added features to this remarkable program and offered improvements.
This entire project -- all 800+ posts -- has strained my credulity to the breaking point. Every time Humanoido is asked to provide concrete evidence for the performance he claims, he resorts either to dodging, referring to someone else's work, inventing new smokescreens in the form of Capitalized Buzzwords, or -- when all else fails -- cheap personal attacks, viz:
Phil, I think you're a bit jealous! Haha! Plus, I always knew you were the one that wears the dress in the family.
Wow, so classy!
Humanoido, I respect the fun you're having with this project and can even endure your need to hype it ad infinitum. (At least you use nice-looking typography.) What I cannot suffer, though, is your endless evasiveness when it comes to posting evidence for the thus-far-unsupported claims you continue making. It's simply not fair to the forum and not fair to Parallax for being such a gracious host to this lengthy thread. But, more importantly, if your claims cannot be substantiated, those more gullible than the few who have challenged you could be misled by allegations regarding the Propeller's capabilities. Worse still, Parallax's credibility could -- by extension -- suffer as a result.
Thanks in advance for understanding -- and responding to -- these concerns,
-Phil
... if your claims cannot be substantiated, those more gullible than the few who have challenged you could be misled...
Back in 1966, Marvin Minsky assigned a homework problem to one his students to be solved over the summer. The problem was to figure out how to make a computer see like a human. Apparently Minsky sincerely believed that human vision was as simple as writing some clever software. In retrospect, such an attitude makes Minsky look either like an idiot or as somebody who was overly optimistic (dare I say arrogant???). Nevertheless, Minsky's employer at the time, MIT, somehow managed to stay afloat despite having such a "confidence" man in their ranks. And, last time I checked, MIT was still standing, even after all these years.
From my perspective, AI people are always making ridiculous statements and unsubstantiated claims. It seems to go with the territory. So I really doubt anyone here expects Humanoido to pull a rabbit out of his (or her?) hat. But who knows? Given another 800000 posts, another 1700000 propeller chips, a few more square miles of breadboards and maybe 3500 more AU's of hookup wire, Humanoido just might surprise us. I'm willing to wait and see.
Comments
Nothing wrong with it just being an art project.
Wow, impressive work Humanoido!
Trust me if you shoot something into outer space, there will be a very bright beacon. Here in Arizona we occasionally see Rockets launched from White Sands Missile Range.
http://www.new-mexico-space-industry.com/assets/images/WSMR-THAAD.jpg
http://wallpapers-diq.net/wallpapers/42/Trail_From_Launch_At_White_Sands_Missile_Range%2C_New_Mexico.jpg
We're talking about a launch site for robotic payloads. Oh Man!
I must say the picture in your first link is the most beautiful Internet Explorer advert I've ever seen!
-Phil
It is possible to make a "node set" with a propeller containing say 28 neurons using scaled integer math. Each "Neural COG" would have to run multiple calculators and signal a storage COG for saving states perhaps in EEPROM. Each neuron could have 4 tokenized "streams of conciousness." How to merge all of that into sentient being is another question.
Apparently there are a lot of collages walking around, chewing gum, making money, driving cars, sipping wine, running for office.
Honestly, I thought the Turning Test was shrugged off years ago. It requires a judge of the test to be fooled - or not. But who is that judge? And who judges the judge to see if the judge is competent to judge? And who judges the judges of the judge, etc, etc, ad infinitum?
I have no idea what the heck Humanoido is doing. But, in any case, it looks like fun.
After reading #837, I'm thinking "Lost in Space".
That does bring to my attention to the bottom line from #837 though ...
"and ongoing Propeller ad Brain development."
I do admire Humanoidio's determination to make something out of all this.
Mind share is always valuable one way or another.
http://forums.parallax.com/showthread.php?124495-Fill-the-Big-Brain&p=1006858&viewfull=1#post1006858
http://forums.parallax.com/showthread.php?125828-Neural-Net-Engine-Abstract&highlight=neuron
As suggested from this thread ... http://forums.parallax.com/showthread.php?t=125627
...there might be a few who wish to explore this concept.
The idea that I have posted is basically a huge 'ring' FIFO shift register with several taps that can either be used as an input or output. When used as an input, external events are allowed to be 'inserted' into the ring creating an 'echo' of the actual events that propagate around the ring. When used as an output, the idea is that you just read the status of the data taps.
To create a flux within the net each bit that is read is stored as the inverse of what was actually read... i.e a '0' is stored as a '1' and a '1' is stored as a '0'. A 'quiet' undisturbed net should have equal number of bits high versus equal number of bits low at any given moment in time. Any external influence on the net should bias the net away from 50/50 and it should remain biased within the ring even after the stimulus has been removed. In this sense the event has been recorded to the 'neural net' cell and new events can use predictive comparisons against previously stored events.
"Beau how do you propose something useful as a standalone device would work with your suggestion?"
http://forums.parallax.com/showthread.php?125627-How-Fast-Can-you-Make-Your-Propeller/page4
That is almost like handing someone in the 1800's an airplane wing and asking what you could possibly use it for.
However, something of great interest to me, in answer to your question would be to setup a 'neural-net-engine'... think if each bit progressed internally on each prop building up a histogram for each pin. outside influences (sensors, even triggers) could affect some bits. Over time many 'F' generations back-n-forth between the props you would have a pattern based on the histogram of an event or series of events. This could be broken down into individual frames, or represent large spans of time. A particular setup like this between two propellers wouldn't need HUGE amounts of memory. This just provides an engine, that perhaps other (read one or several) propellers could tie into the 'neural-net-engine' and contribute.
Think of a 'Chair', a simple chair. You had to be told that it was a chair right, but now subconsciously, you know when you see a chair, only because you have identified it with some kind of reference 'data' or histogram. Now imagine several chairs, where each time the histogram builds upon the previous data, and only the common mode information that is pertinent remains.
Beau Schwabe
Parallax Semiconductor IC Layout Engineer
I want to see some big brain rocket experiments...
When you have a demo instead of a bunch of links I will be impressed.
Until then it is just an art project with which you woo the all the Dr. Jim suckers.
Yet, when pressed about actual running software, he quotes an abstract concept by Beau that may never have been fleshed out. I'm all for fanciful art installations, and it's fun to construct "castles in the sky"; but when claims are made of actual performance, I think we deserve to see details that can be replicated.
Where is the Propeller code for these "neurons", Humanoido? "Enquiring Minds Want to Know." TM
Thanks,
-Phil
Ttailspin, as you can see in the photo, Phil and his volunteer Propeller team are working around the clock on this homebuilt rocket spacecraft to launch the brain. As Phil said, "see that big black hole? That's where you stick your brain."
From left to right: Phil attaching his one of his many stars. Jazzed: a little hammered. ElectricAye crow bar adjusting the nose cone and inscribing Calculus & differential flight equations. Mike G in the suit with pre flight jitters holding the remote. Fido looks on.
Who was taking/painting the picture?
I think this quote puts much of what is needed from the big brain into perspective... The brain has to learn much, so it can do less...
Indeed. And for some theory on how to do that, here are some nice papers:
http://www.cnbc.cmu.edu/ibsc/papers/RogersMcC.pdf
http://www.cnbc.cmu.edu/~jlm/papers/McCRogers03.pdf
This builds on other work done by Rumelhart, McClelland, and the PDP research group.
http://www.cnbc.cmu.edu/~jlm/papers/PDP/
http://en.wikipedia.org/wiki/Connectionism
It's Ok to want a respectable and functional brain project.
Yet another reason to bring back the Sandbox.
What does this have to do with robotics? (In the hardware realm which this sub forum is based on).
The Big Brain is being prepped as a robot. It already has a Brain Stem (see index) dedicated to motion control and will be outfitted with servos, mechanics and various limbs.
It's a constituent part of a robot, a robot brain. A refined condensed version will be used with humanoid robots. Smaller versions of the Big Brain are being prepared for smaller robots. This is a large ongoing project with puzzle pieces that fit together over time but won't be fleshed out until completion.
The taxonomic hierarchy of the type used by Collins and Quillian in their model of the organization of knowledge in memory appears very similar to the programmed behaviors of Lerner, which is learning code already completed and running on the Propeller.
Demo?
Phil Pilgrim, thanks I can see you're very interested and feel like an impatient child waiting for Santa's gifts on Christmas eve... Big Brain says, "Code is expected to post upon project completion. Have patience."
The taxonomic hierarchy of the type used by Collins and Quillian in their model of the organization of knowledge in memory appears very similar to the programmed behaviors of Lerner, which is learning code already completed and running on the Propeller.
Download the code in the Big Brain source found below:
The Big Brain Thread, Page 27, post number 523, Machine Learning with Lerner - Dave Hein fills a small brain! Dave has since added features to this remarkable program and offered improvements.
or
Wow, so classy!
Humanoido, I respect the fun you're having with this project and can even endure your need to hype it ad infinitum. (At least you use nice-looking typography.) What I cannot suffer, though, is your endless evasiveness when it comes to posting evidence for the thus-far-unsupported claims you continue making. It's simply not fair to the forum and not fair to Parallax for being such a gracious host to this lengthy thread. But, more importantly, if your claims cannot be substantiated, those more gullible than the few who have challenged you could be misled by allegations regarding the Propeller's capabilities. Worse still, Parallax's credibility could -- by extension -- suffer as a result.
Thanks in advance for understanding -- and responding to -- these concerns,
-Phil
Back in 1966, Marvin Minsky assigned a homework problem to one his students to be solved over the summer. The problem was to figure out how to make a computer see like a human. Apparently Minsky sincerely believed that human vision was as simple as writing some clever software. In retrospect, such an attitude makes Minsky look either like an idiot or as somebody who was overly optimistic (dare I say arrogant???). Nevertheless, Minsky's employer at the time, MIT, somehow managed to stay afloat despite having such a "confidence" man in their ranks. And, last time I checked, MIT was still standing, even after all these years.
From my perspective, AI people are always making ridiculous statements and unsubstantiated claims. It seems to go with the territory. So I really doubt anyone here expects Humanoido to pull a rabbit out of his (or her?) hat. But who knows? Given another 800000 posts, another 1700000 propeller chips, a few more square miles of breadboards and maybe 3500 more AU's of hookup wire, Humanoido just might surprise us. I'm willing to wait and see.