Slashdot Mirror

Number one, would we know such an intelligence if we saw it? Furthermore, could we determine what such an intelligence was thinking if we could? I think we could deduce that something was behaving in an intelligent manner, if we applied the right tools to the purpose. The danger would be in not knowing if the object we are studying feels that such testing is against its interests and acts to stop the testing. For example, a theory of emergent behavior within large groups of people (think large bureaucracies or societal constructs) might indicate the possibility of a "group" or "hive" mind arising from interactions between the individuals involved, that is both of the individuals yet outside of it (same as Mind is to Neurons). The output of such a "mind" might seem to be intelligent, but is there any way to actually know what it is thinking, or how it is communicating? Can a neuron ever know of the human, or brain, or mind? What would we (as a human) do if a neuron suddenly could understand? Is it in any way possible such a group mind would act in the same manner? Would we understand it if it did? Can we assume that such a thing isn't happenning already in our increasingly connected and interacting world?

Secondly, I think the other problem with building such an artificial mind is that of design and construction. Interestingly, we likely have both at hand. For design, I subscribe to the view that the mind (or at the very least, the cerebrum) is nothing more than a pattern recording and playback machine, as detailed by Jeff Hawkins in his book On Intelligence. I am pretty certain that this idea is spot-on, and is something that should be investigated much further. As for construction, the design of Dr. Hugo de Garis's CAM-Brain Machine (CBM), as realized by Genobyte, seems to be the approach to use to build a system similar to what is described in On Intelligence. These machines were actually built, shipped, and used in a few research institutions around the world. Whether they still exist or not, or are buried in a back room, is anyone's guess. The fact is that they aren't a standard design for a computer, and furthermore they utilized Xilinx FPGAs that isn't manufactured anymore (whether a similar machine could be built using a different Xilinx FPGA is another matter), leads me to wonder what will happen to these machines as they end their useful lives and/or have hardware failures. Also, it doesn't appear that Genobyte is in business anymore, though their website still maintains "ghostship" status.

Maybe I am reading too much into either of these ideas? Maybe both are a bunch of hooey (indeed, the whole CAM-BRAIN machine thing is something that I am not sure whether to completely believe or not - I seem to remember a /. article a long time ago in which another company linked to this - STARLABS - was seen to be a hoax or something?). Even so, the ideas seem sound, even if the implementations don't exist in fact (although, all the research I have done seems to indicate that these systems do in fact exist).

Sounds like a cross between what Jeff Hawkins described in On Intelligence, and the FPGA evolvable hardware of the CAM-Brain Machine project...

A long time ago (or so it feels), a company existed which seemed - both at the time and even now - to be one of those "fly-by-night" dot com scams that was flying around near the time of the bubble. It seemed like a pump-and-dump scheme, a scam to get peoples money and run with it. At the same time, it seemed to hold out promise for the possibility of AI, and if not that, than at the very least the possibility of a radically new type of computer architecture, on par with that of Hillis' Connection Machine.

The company was (is?) called Genobyte, and the system they (supposedly?) developed was the CAM-Brain system. Relying on a certain model Xilinx FPGA (the Xilinx XC6264BG560 - one which, as I have seen noted, either on the website or in other papers on the net - is likely no longer available, and that they bought up the last of them for their efforts), the system essentially combines the ability to "program" an FPGA by using cellular-automata software to "evolve" neural networks, the final emergent design which is then loaded into the CBM, hopefully to solve useful problems. Essentially, what was developed was a hardware-based neural network processing system, the design of which was evolved using a cellular automata system, to solve problems.

But, as I have noted, what they made seemed (seems?) akin to "vaporware". Sure, they have pictures of the machine (and I must admit I find the computer to be one of the best looking designs for a computer since the Cray 2), and the ideas and theory seem sound. Supposedly, they installed at least two machines: one in November 1999 at Kansai Science City, Kyoto, Japan (ATR Human Information Processing Research Laboratories), and one in December 1999 at Ieper, Belgium (Flanders Language Valley). If this is all true, then these machines must actually exist somewhere, right?

What is the truth on this company, and the people behind it? Did they actually create what seems, at least on the surface, to possibly be a real advance in computational hardware? Or, is it likely to be another scam? Their website hasn't been updated in years, but somebody is paying the bills to their registrar and/or hosting provider (or whatever they are using - the electricity is still flowing to host their site). So - anybody have any idea what this company is, or is not? What about the people behind it?

Is it a scam, or a "revolutionary" real product?

A long time ago (or so it feels), a company existed which seemed - both at the time and even now - to be one of those "fly-by-night" dot com scams that was flying around near the time of the bubble. It seemed like a pump-and-dump scheme, a scam to get peoples money and run with it. At the same time, it seemed to hold out promise for the possibility of AI, and if not that, than at the very least the possibility of a radically new type of computer architecture, on par with that of Hillis' Connection Machine.

The company was (is?) called Genobyte, and the system they (supposedly?) developed was the CAM-Brain system. Relying on a certain model Xilinx FPGA (the Xilinx XC6264BG560 - one which, as I have seen noted, either on the website or in other papers on the net - is likely no longer available, and that they bought up the last of them for their efforts), the system essentially combines the ability to "program" an FPGA by using cellular-automata software to "evolve" neural networks, the final emergent design which is then loaded into the CBM, hopefully to solve useful problems. Essentially, what was developed was a hardware-based neural network processing system, the design of which was evolved using a cellular automata system, to solve problems.

But, as I have noted, what they made seemed (seems?) akin to "vaporware". Sure, they have pictures of the machine (and I must admit I find the computer to be one of the best looking designs for a computer since the Cray 2), and the ideas and theory seem sound. Supposedly, they installed at least two machines: one in November 1999 at Kansai Science City, Kyoto, Japan (ATR Human Information Processing Research Laboratories), and one in December 1999 at Ieper, Belgium (Flanders Language Valley). If this is all true, then these machines must actually exist somewhere, right?

What is the truth on this company, and the people behind it? Did they actually create what seems, at least on the surface, to possibly be a real advance in computational hardware? Or, is it likely to be another scam? Their website hasn't been updated in years, but somebody is paying the bills to their registrar and/or hosting provider (or whatever they are using - the electricity is still flowing to host their site). So - anybody have any idea what this company is, or is not? What about the people behind it?

Is it a scam, or a "revolutionary" real product?

Why am I so certain that we are close? We have at our hands numerous pieces of technology and knowledge, some fairly mature, others fairly new, which, taken together, can be seen to represent a possible whole greater than the sum of the individual pieces. The key pieces (and there are many other important pieces, but not as key, such as recent advances in growing neural nets using real neurons, and growing neurons on a silicon substrate, among others) are:

• Network theory and applications
• Neural network modeling (and associated neuron modeling)
• Artificial Life (CAs and evolution modeling, such as being discussed in the article)
• Theory and Evidence on intelligence (and likely consciousness/sentience) arising from pattern matching
• The Genobyte CAM Brain Machine

Network theory and Neural network modeling has yielded, over the years, a ton of insight into how brains work (as well as a lot of data in how they likely don't work). Such modeling has taught us a lot in this area, but has yielded little to nothing on what we call consciousness or intelligence. These systems are forced to be small (relative to our brain - or even the brains of lesser creatures) due to the fact that currently the only practical method to simulate such systems is via software - simulating an analog, non-discrete, networked, and distributed system on a digital, discrete, serial computer. These small systems have yielded some very useful and practical technology (improved OCR and better credit fraud checking software, for instance), but none has created a true intelligence. The majority of the limits are imposed by the fact that you have to simulate a large networked system, where each node consumes many bytes of memory, along with many more bytes to describe the interconnects, and still more bytes for the software to simulate all of that - that ultimately you are left with a very hairy problem that a serial computer takes time to churn through. While a network with trillions of nodes could in theory be simulated in software, given enough hardware, it currently isn't something that can be done practically. Given these constraints, researchers have had to be content with studying smaller networks, hoping to interconnect them in some way to make a larger, more powerful whole. These efforts have yielded some results, but as a whole these "hand-tuned and taught" networks are not likely to be the building blocks we hope them to be. Ultimately, it will take a very large neural network to be built to make an intelligent system.

Some researchers have speculated that a simulation may limit these systems precisely because they are not "non-discrete", and that it is the ability for a brain to take in multiple inputs all at once and process them all at the same time to yield outputs which is responsible for what we call "intelligent behavior". Whether this is true or not, we don't know, simply because we haven't been able to build large scale non-discrete processing neural networks. On the other side of the coin, it has been speculated by researchers that the brain simulates serial computation from the inherent pattern matching ability of the neural network which makes up the brain. This is a very interesting possibility, one that argues, in a sense, that a neural network can become and simulate a UTM for serial computation (mathematics and logic). This isn't surprising - if we can simulate a neural network using a UTM, why can'

GenoByte has found a more novel use for FPGAs, which they call "evolvable hardware." Much like our own brains neural networks on the FPGAs reconfigure the way they interconnect on the fly; commonly used paths are reenforced while less frequently used ones atrophy.

Here are some cool pictures:

The CAM-BRAIN machine, a big box full of FPGA boards: http://www.genobyte.com/images/machine.jpg

Neural network layout for the XC6216 FPGA: http://www.genobyte.com/images/chip.JPG

All in all this approach is substantially faster than modelling large neural networks on a general purpose processor. In the GenoByte approach, the neural network is implemented as physical circuits.

GenoByte has found a more novel use for FPGAs, which they call "evolvable hardware." Much like our own brains neural networks on the FPGAs reconfigure the way they interconnect on the fly; commonly used paths are reenforced while less frequently used ones atrophy.

Here are some cool pictures:

The CAM-BRAIN machine, a big box full of FPGA boards: http://www.genobyte.com/images/machine.jpg

Neural network layout for the XC6216 FPGA: http://www.genobyte.com/images/chip.JPG

All in all this approach is substantially faster than modelling large neural networks on a general purpose processor. In the GenoByte approach, the neural network is implemented as physical circuits.

GenoByte has found a more novel use for FPGAs, which they call "evolvable hardware." Much like our own brains neural networks on the FPGAs reconfigure the way they interconnect on the fly; commonly used paths are reenforced while less frequently used ones atrophy.

Here are some cool pictures:

The CAM-BRAIN machine, a big box full of FPGA boards: http://www.genobyte.com/images/machine.jpg

Neural network layout for the XC6216 FPGA: http://www.genobyte.com/images/chip.JPG

All in all this approach is substantially faster than modelling large neural networks on a general purpose processor. In the GenoByte approach, the neural network is implemented as physical circuits.

Look at my other post in this series. There is a project to evolve 32,000 neural net modules containing up to about 1000 neurons each in an electronic brain of a robotic cat. The early stages are emulated in software, including the motions of the cat, gravity, limits of the motion. Later the evolved nets are transferred to FPGA's and placed in a physical robot to contiue the evolution. Here is the link again. LINK

Exactly what you ask for, but in a cat.

Rumor has it they plan on turing it into a consumer robot sometime in the next year or so.

Exactly what you ask for, but in a cat.

Rumor has it they plan on turing it into a consumer robot sometime in the next year or so.

Here are a few links that you might find interesting. Xrays of the dog. http://www.nnc.ne.jp/~as212/aibo/x-p.html Aibo disassembled. http://www.aibosite.com/index-s.html Aibo taken apart. (in japanese) http://web.kyoto-inet.or.j p/people/nktks/TEMP1/menu1.htm Robotic Cat - (evolved neural net) http://www.genobyte.com/robokoneko.html