Dr. Jim's code is very tight. It needs more than 32k, but can fit on 4-16 MB easily, 2 if he squeezes very tight and holds his breath while using a shoe-horn. He is using the Propeller for the cogs. Each cog is fast enough to do what he needs it to do. It is all about parallel design. He comes from the mainframe world and is building a sort of "mainframe" architecture for this project.
Dr. Gouge and Mr. Allred,
Very boring clips. I didn't hear anything that was grossly wrong, but the general impression was that of a lot of not very well connected, but well known information with nothing of any depth about the supposed subject of the talk. There was nothing I heard that wasn't written about in books and magazines years and years ago, and discussed much better.
I've worked in the area of "holographic" machine memory in the past and the same kind of pitches were used then. Technology is much better now and we were working with much slower and smaller computer systems than the expanded Propeller system you're talking about. Unfortunately, real machine intelligence is much much more complicated than undifferentiated neuronal networks and we're only just beginning to understand some of the deep organizational structure involved in the brain.
You may have some great ideas, but your skills at presenting them need work. Some successful bright people with great ideas find others with the technical presentation skills and provide them with the information needed to do their job. Right now unfortunately you and Mark sound like snake oil salesmen and these video clips don't help. Saying "it ain't so" and "just wait, you'll see" doesn't make it different. It would be a terrible waste to have good ideas and just a lousy presentation.
Mike, I respect your opinion, but we are working on a shoestring budget. We are doing the best we can. I'm sorry if it is not good enough for you.
We have a following that thinks we are on to something, and it is for those people we will continue to do our best to convey the technology as best we can within our budget.
can't see how you can be on shoe string budget if people are buying your stuff. the cost to build a 16mb memmory expander is a tiny fraction of what you charge(not saying making a profit is bad). How big is your following and are they also waiting to see the finished product before buying your stuff?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us and propmod_1x1 are in stock. Only $30. PCB available for $5
Want to make projects and have Gadget Gangster sell them for you? propmod-us_ps_sd and propmod-1x1 are now available for use in your Gadget Gangster Projects.
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
Our numbers are confidential, however, our following are buying without seeing the finished product. They are taking a risk, to be sure, but the great thing about it is, they have enough confidence in our approach to purchase our products so they can become familiar with them before the project is finalized. These are our early adopters, and they will have a head start when the wider public starts accepting our approach.
The wider public will come on board when they see machine intelligence on the propeller, first with acceptance of voice recognition as we prove out the concept and show video clips of it on our blog, then as we add binocular vision and response, tactile input and response, ultrasound object detection and response, motor functions, etc. When this is all working, we will have a large audience indeed.
And we are on the cusp of all of this. You are entitled to your opinion, but mine is that we will succeed.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us and propmod_1x1 are in stock. Only $30. PCB available for $5
Want to make projects and have Gadget Gangster sell them for you? propmod-us_ps_sd and propmod-1x1 are now available for use in your Gadget Gangster Projects.
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
Dr. Jim is out right now mailing the latest ERAM board we sold, so we are having a measure of success. The people on board love what we are doing, and they are thanking us for bringing the hobbyist along with us. That has been some of our feedback so far.
mallred said...
[noparse][[/noparse] In response to Mike's suggestion that they get a good technical publicist ]
Mike, I respect your opinion, but we are working on a shoestring budget. We are doing the best we can. I'm sorry if it is not good enough for you.
Lot's of people work on a shoe string budget. I'm sure that you have a university or college near by with lots of students willing to work for cheap while still having the technical knowledge to make a good presentation.
As a side note, I was going to suggest a couple of places around where you are that could supply some students. Unfortunately, your website doesn't list any street address or even a telephone. I have to ask: what reputable company only allows communication via the web? A PO box only costs about $60 a year, and a telephone number isn't much more.
Edit: Technically, I'm certain that any business (at least in California) has to have a registered address and possibly a phone number on file at a public registry. So, therefore, you have an address that is floating around out there somewhere. Why not share it? I do hope that you incorporated to prevent a lawsuit from taking personal property.
@SRLM: yes indeedy. The mail drop and phone are generally lessons #1 and #2 in How To Appear To Be A Real Company. With pay as you go cell phones there isn't really any excuse.
There is certainly nothing wrong with coming home from your real job to do your own designs, take web orders, and put stuff together in your garage for sale by mail-order; several people here are very openly doing that and sometimes they offer real bargains, as when they've done the SMT for us or sourced components that are a lot easier to get in quantity. And let's face it, just about everything that exists so far for the Propeller was built first in someone's garage.
But there is something wrong with pretending to be something you're not. Designing a memory expansion is not special; there's been a lot of discussion here about how to do it. Popping in to claim that your memory expansion is the bestest thing ever when you aren't even aware of those discussions impresses no one. Designing an operating system is not special; there are half a dozen similar projects in the works, some with significant progress. Again, the relative merits have been discussed extensively and all of them will have a place, because they are aimed at different goals. Designing hardware I/O controls is not special; hell that's Parallax's whole business model and half of the contents of the obex.
So we have items A, B, and C, are reasonable enough to assume they have emerged from your garage of businessdom in relatively non-vapor form. Then we have item D, wherein you claim that items A, B, and C will allow you, with the aid of a few eight dollar chips, to solve a problem that has continuously beaten very smart people who are paid well to spend all of their time going at it with supercomputers and hundred thousand dollar robots. And that furthermore the imminence of item D implies that items A, B, and C must be highly desireable and worth considerably more money than similar products offered by other garage dwellers.
Furthermore, when pressed about item D, you reveal that you are ignorant of what is actually known about the biological solutions that exist, you're ignorant of current research in the field, and you're even ignorant of a fairly prominent philosophical movement which is OBSESSED with item D that has generated several bestselling books and thousands of blog posts. All right, not a lot of people outside of the Singularity community have heard of Eliezer Yudkowsky, but surely you've heard of Ray Kurzweil? And the DARPA challenge? You do realize the people who won the DARPA million did it by doing almost exactly what you claim you will be doing?
I am someone who has thought about AI a lot. I have been convinced since I was a child -- in the 1970's -- that strong AI is possible and will eventually be created. I also believe it will be self-programming through experience and much of the reason it hasn't been done yet is that people in the field are too focused on specific results instead of trying to get results to emerge from chaotic interaction.
But I do not believe you are going to get a machine that learns English on its own and asks sensible questions from a platform with a few propellers and 16 MB of RAM. I don't believe it because if it were that easy, it would have been done already on a more expensive and powerful platform. The prop is very clever, but we're not here because it's the fastest thing out there; a lot of DSP chips could run circles around it, 8 cogs or no, and a bottom of the line desktop PC also has more power even with the context switching overhead. And that's not even to mention things like GPU cards and the Cell processor.
So you have a memory expansion and some I/O routines, very nice, they might be useful in a lot of ways. But please forgive us if we are not impressed when you claim that they are our chance to get in on the ground floor for owning our own Lieutenant Data. Or even Skynet.
@Mark, do not feel bad about some of the comments that are being posted here. I do not know if you were following this site when Gadget Gangster was starting up his business, there were posters here that were making all kinds of accusations. I think those posters have all moved on, but wait, they are now targeting you.
Having said the above, the Japanese are well ahead of you in this field, they already have robots (machines) that can do just about everything that has been talked about here, and soon enough they will be marketing here. So, time is not on your side, you may have an excellent product, but what good is that if they beat you on your own turf.
You just have to stick with it, most of these skeptics are not going to purchase your product anyway, so do not fall for there baiting schemes.
Now, that the SX chip is EOL I will have to start thinking more about the Prop chip. Since you guys are doing the voice, I may have to go for the replication of striated muscle function.
That's cool! We need striated musculature in our bots!
Like I said, what Dr. Gouge is attempting may have been done before a different way. That's ok by us. We are attempting to do it our own way. This should not be the only way, but we feel it is a good way, so we are going for it. I know it is controversial because almost no one has figured out how to do it well. We are taking our stab at it. Mabe one of you will have an even better idea than what we develop. That's ok too. We would love to see it. But we are proceeding with our project just the same.
I have seen some of the things the Japanese have done. Some of their work is quite impressive, but still appears to be programmed. I saw a model that cost $50,000 for a robot and was nothing more than some fancy hardware and software, and a remote mic where a person sitting across the room could hear and respond to people's questions. That, I believe we can all agree, is fraud.
I think that he is attempting to do it on the Propeller BECAUSE it is a small chip. Doing on the PC is hard, yes, but doing it on a MICROCONTROLLER?!?!?! Definitly worthy of recognition. He is trying to do this in a way that has never been done before. He is trying to create a rough AI that is affordible and accessible to the hobbiest. I hope that he achieves what he has set out to do.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Toys are microcontroled.
Robots are microcontroled.
I am microcontroled.
If it's not Parallax then don't even bother. :-) ·
Mini-Din/PS2 connectors are for sale! 5 for $1! PM me if you wish to make an order. Cheap·shipping unless specified!··········150 left!!··
Post Edited (microcontrolled) : 7/31/2009 9:58:43 PM GMT
@microcontrolled & mallred -- When you are approaching a problem that everyone agrees will require massive computational power, you do not throw a microcontroller at it.· If you need it to have a mobile sensory platform to collect experiences there are these things called radio waves that can be used to link it to the energy-hungry brain.· Nowadays you can do such a link with about a hundred bucks worth of wifi modems.· You could even use props on the robot with a cheap ethernet adapter.
If you don't want the complexity of·a multi-tasking operating system·you buy a basic PC, reformat the hard drive, and install FreeDOS.· You now have a single tasking system that will run rings around a Propeller and also supports lots of·fancy PC-based I/O devices such as hard disks and cameras.·The Propeller is not more powerful than a PC.· I have programmed both, and it's just true.··It's also not harder to·program if you know what you're doing; you can build your own multitasking state machine engine·for a lot less effort than the complexity of rigging up megabytes of RAM to a propeller.· The PC will use a lot more power but it doesn't even cost that much more when you factor in the·available I/O buses and memory.· We are not here·in Propellerland·because the Prop is more powerful than a PC, we are here because the Prop can generate video in software (which in turn has more to do with deterministic timing than superfast timing), its I/O pins are general purpose and can·source/sink 40 mA, and it runs on AA batteries.
Let me be perfectly clear on this; I have done this kind of programming.· For awhile I had my eye on the DARPA challenge.· (That lasted until I realized that if I succeeded, building SkyNet and giving it to the very people who were most likely to turn it into SkyNet was probably a bad idea.)· I agree in principle with many of Dr. Jim's stated ideas.· But I am not here because I think the Propeller is the key to building a Seed AI.· I'm here because the Propeller can really reduce the parts count and design complexity of embedded devices.· It occupies an otherwise very empty corner of the price / performance map in that respect.· It is not, however, supercomputer material, and I strongly suspect Chip Gracey would agree with me on that.·
There are other things out there that are low-cost candidates for this project, such as GPU cards and the Cell processor in the Sony PS3, both of which can run circles around an entire farm of Propellers.· With the Prop and the memory board I will grant that you might get something that can find its electrical socket when the power runs low.· But the very idea that it will learn natural languages to communicate with humans is not just laughable, it's stupid.
The reason I chimed in on this is that I detect an unfortunate pattern here.· The business operated out of a garage pretending to be bigger than it is is no rarity and not much source of shame, though I admit I have more respect for people who are more up-front about their real capabilities.· It's the hyperbolic claims, the selling things sight unseen based on these hyperbolic claims and perhaps a cult of personality, that worry me.· This reminds me very much of a scam that was pulled in the late 1970's when semiconductor RAM was hyper-expensive, lots of computers still used core RAM that was created by manually threading wires through teeny ferrite donuts to create a bit,·and there was a technology called "bubble memory" that was supposed to be the Next Big Thing.
Several companies jumped on the bubble memory bandwagon, attracting investors by promising that their bubble-memory based tech would create exactly the same things Dr. Jim is promising -- computers that would talk with you, play chess and beat you (that hadn't been done for real yet), even control android robots.· Bear in mind that in that day a megabyte of RAM was the kind of extravagance only a government or multinational corporation could afford, so this convinced a lot of people.· There were VHS videos and public demonstrations where a "bubble memory controlled" robot would play chess with audience members.
Turned out it was all a scam; the promising demos, including the chess-playing android robot,·had been faked (it was remote controlled by an actual human chess master), and the investment money disappeared along with the founders when the lie got too hard to hide.
I hope that Dr. Jim is just naive about the possibilities for his self-educating network; it's an approach lots of other people are working on with far more capable hardware though.· But the possibility that he has plugged "parallax propeller" into an old, old script in place of "bubble memory" would be a much uglier scenario.
First, we have to get one thing straight. We do not require massive computational power. We require a massively parallel system, which is why we need the cogs, which is why we need multiple Prop Proto boards. Each cog is fast enough to do each function necessary, but you have to overcome this idea that massive computational power is needed. That is NOT the case with our approach.
Dr. Gouge wrote the telemetry for the LEM (lunar lander) during the Apollo missions. They had to page up 4k at a time due to the restrictions of memory at the time. It was a mainframe world back then. As a result, he writes very tight code and thinks in terms of mainframe architectures.
Since he is able to create his own platform his own way, he has plenty of power in each cog to do what he needs to do. He simply needs many cogs to do it. Again, get away from PC architectures and look more to mainframe technologies and you will understand our approach a little better.
We are not doing any kind of remote brain using RF or microwave, etc. because it is far too slow to do what we need it to do. We have chosen static expanded RAM to become the neuronal structures because they are very fast and will be able to react quick enough for Dr. Gouge's machine intelligence algorithms. And as Dr. Gouge explained in his videos, expanded RAM was necessary, or this project would not be able to get off the ground.
If you wanted to use a PC chip approach, you would need 32 to 64 computers to work out the same approach. The cost would be prohibitive for the average hobbyist and far too bulky for our purposes. We need to be mobile. As it is, you will end up with a "cube" of processors and memory that will need to be installed onto the robot.
You are absolutely correct about Propeller pricing. This is a major factor why we chose it and why we can use multiples of the Prop Proto board for the COGs we need for our project. Four or eight boards is within the price range of anyone desiring to do this project on a kitchen table.
But again, we are not building a Supercomputer. The architecture is more like a mainframe. Supercomputers need to be super fast. We do not need that. What we need is massively parallel. So again, think mainframe, not supercomputer.
The other hardware that you mention, such as GPUs or PS3s would also be prohibitive cost wise for the same reason, you need multiples of these to do the trick, and you do not want to mess with multithreading etc., way too much overhead. COGs work perfectly and at the right price for our intended architecture. Believe me, we thought this through when we chose the Propeller.
About learning natural languages...just wait and see. You may be surprised.
While we may be working out of the garage (spare room that we turned into a lab), we are not liars or scam artists. You will believe what you want to believe and that's fine. But keep an open mind.
From many of these posts, I still see too many people thinking the wrong way in a PC world where faster is better. We don't need super fast processors. Just fast enough processors and a lot of them. Start thinking about mainframe architectures and you will be thinking much closer along the lines of what we are doing.
John Abshier said...
I have a question that won't go away. Why is Dr. Jim using a Propeller? I think the final design requires 3 or 4 Propellers. If you need lots of memory and an OS why not use a PC? If you give me a system with AI, vision, speech recoginition, speech output, I will build a robot that can carry a laptop.
John Abshier
This is where so many people go wrong. Throwing a PC at the problem is not always the best way to do something. Back when I was younger, we didn't have all of these powerful microcontrollers to run things. We had to figure out clever ways to do things. Look at the robots from hundreds of years ago - they didn't even have electricity, yet they managed!
A PC with a 1TB hard drive and 8G of RAM is a sure way to guarantee that the bloat will eat up your project. And using MS-DOS or (worse), Windows? Don't even get me started! The power requirements of a PC are horrendous. All of that bloat requires more battery power, which requires a more powerful robot to carry it, which requires a bigger battery to move that heavier robot. Going with a clean slate, and starting with a processor that was designed for parallel processing, instead of having it tacked on, along with tools that are designed to support it, and not to sell bloated consumer code to dullards, seems like the more sensible way to go.
edit: I started this post before the above post was posted. Looks like we're thinking along the same lines. Think Different, people! Get out of your mental ruts, and the world will open up for you!
Please stop trying to impress me with references to mainframe computers. When in college I was taught FORTRAN on a DEC SYSTEM 10, which if I'm not mistaken is a mainframe computer. I have also done real work on 70's-era 8 bit PC's, and more recently on similarly powered equipment like the 8051. I know all about tight code and doing it with integer math in assembly language. I am 45 years old and I have all those T-shirts in my closet.
Massively parallel systems are about computational power. Saying they are not about that is like saying Beyonce is not about sex. The Prop has 8 cogs because it would have to run at 640 MHz to give equivalent computational power without the parallelism, and it's a very conservative design for these days that keeps the frequencies out of the UHF range. A similar non-multiprocessor solution running at 640 MHz should be able to do what the Prop does at 80MHz. In reality, there are other factors, such as context switching and the fact that cogs are really, really fast because of their 2K cog memory limit which mean the other systme would have to be a bit faster to emulate a Prop, but when common PC's are running 2 GHz processors and megabyte on-chip RAM cache, it's very silly to think a propeller can do something non timing-dependent that a PC can't.
You do not need 32 to 64 PC's to do what one propeller can do; you need one. The one thing the Prop can do with absolute constancy the PC can't is generate timing and waveforms. The PC will be as fast in processing but not as reliable on a microsecond level. This is of absolutely no consequence in the kind of software you are talking about. It's very important if you want to generate video or RS232 signals in software, but irrelevant in neural network simulation.
I'd be interested in an actual explanation (with actual numbers and examples) of why RF remoting isn't practical, when RF cameras delivering realtime video over wifi are under $200. I am pretty familiar with the entire range of sensors you might be thinking of installing on your android, and especially considering that your financing seems to be on the same level as mine, I can't imagine any combination you could afford that would saturate a wifi link.
What you don't seem to understand is that the propeller is, by modern CPU standards, SLOW. Cogs do not change this. CMOS gates use energy when they change state, so 8 cogs at 80 MHZ will use the same energy as one similar single-cog CPU at 640 MHz. That is still slow by modern standards, and there are many DSP chips that blow smoke rings around the Prop while being designed for things like processing video.
As for your product learning natural human languages -- I will wait, but I dont' think I will be surprised. I will be the first to salute and offer apologies if I'm proven wrong.
localroger said...
when RF cameras delivering realtime video over wifi are under $200.
Not to hijack the thread, but this is a significant bugbear of mine. IP cameras over WIFI are *not* real time. There is significant latency which poses a number of challenges to the designers of CCTV systems with IP cameras. They are also unreliable and can be easily interfered with. Both less than desirable qualities (slow and unreliable) when dealing with CCTV.
I'd sure as heck hate my Robot to be slow and laggy when it's trying to pickup a crystal scotch glass.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
lt's not particularly silly, is it?
I respectfully disagree with almost everything you had to say.
Anything I could say, you will disregard, and likewise, most of what you had to say is not applicable to our machine intelligence technology, and many of your facts are wrong, so we are at an impass, with due respect. For example, DEC never made a mainframe. They made MINI computers and MIDI icomputers, but never mainframes. A DEC PDP-8 MINI computer had a fortran compiler as well as cobol, et. al. That does not make a PDP-8 a mainframe. PDP-10s and 11s were mid-range computational platforms, or MIDI computers. If you want to talk about mainframes, there were only a few. Mostly, Univac, IBM, and Control Data Corp. These were the major mainframe suppliers of the time.
Our technology is not based on any neural-net paradigm, so the computational power needed for those is not required for our architecture. We are more interested in the parallelism, not having to do context switching. We don't need time-slicing, but rather true multiprocessing, which the Propeller provides nicely.
If you are looking for an example, look to the biological model, not previously tried technologies based upon neural nets, fuzzy logic, etc. Those have all been tried for decades with limited/questionable results. As an example, there is one question to answer. Where is the truely intelligent robot? It has not yet been made.
Our technology promises to provide a major step in the direction of autonomous, intelligent machines and provides an excellent, low-cost platform for future research. Isn't that what it's all about?
I like multi prop designs and is why I made a board that holds 8 In sync for 200 but I know from experience it is relatively easy to write an os to take 1 processes to emulate many.
I can see space being an issue. A small pc is 4"x5" my brainboard is 1"x7l but since you are using dev boards and dip ram I assume that is not a care of yours.
Cost my brain board is $200 pc is about the same.
I hope it works. Your system is to big but I would be willing to work with you or buy a license to make a smaller version for myself
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us and propmod_1x1 are in stock. Only $30. PCB available for $5
Want to make projects and have Gadget Gangster sell them for you? propmod-us_ps_sd and propmod-1x1 are now available for use in your Gadget Gangster Projects.
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
Dr. Jim knows the details of our architecture way better than I do, but I can see the advantage of having 8 propellers on one board (forgive me if I am stating this wrong). Perhaps it would be good to collaborate with Jim on merging your multi-prop board with our expanded memory and that way we would have 64 COGs at once on a much smaller physical footprint.
I know Dr. Jim wants to use the serial mouse to talk to multiple Prop Proto boards. He has reasons for using multiple boards, for use of additional I/O pins (which you could probably access via your 8 chip system just as well) but also for USB devices for thumb drives, etc. Even if we could not use it for all the boards we might need, it would make a superb "Master" board if we could work it into our OS. It's still probably worth asking him about it though. You might shoot him an email at support@machineinteltech.com. I'll make sure he gets it.
Also, if you know where we could get parts to reduce the cost of our ERAM boards, we would be interested in lowering prices so more people could get involved.
The PDP-10 was a mainframe by any standard of the day. It had a 36-bit word with complex 36-bit instructions including a hardware stack, a full register set, hardware floating point, a large memory with hardware paging assistance (virtual memory), an operating system that had high level features not duplicated to this day that provided multi-user timesharing, general multitasking, Internet services like Telnet, FTP, full e-mail.
Post Edited (Mike Green) : 8/1/2009 3:39:23 AM GMT
apparently you never heard of a DEC VAX; a commonly
used mainframe. i know because i programmed on one
twenty years ago.
you and your friend jim seem to ignore common sense.
but what really bothers me (and apparently a lot people
on this forum) is you write a lot of words without saying
anything at all. i find hard to believe anything you say,
especially that you sold ANY of your wildly overpriced
memory boards.
tired of all this
blake
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"Can't sleep, clown will eat me."
Bart Simpson
Oh, lawsy, lawsy! This is like watching a Greek tragedy unfold, complete with hubris and even a chorus (i.e. this forum). The hubris is pervasive, from the constant touting of the title "Dr." (doctor of what, exactly, and from where?) to the self-aggrandizing puffery and hand-waving that we've been subjected to in multiple posts and in the blog videos. I hope the Fates are kinder to Dr. Gouge and his close family business than they would be in a Greek tragedy. I really do. Otherwise this will be like watching a train wreck in slow motion, and that would be very sad, indeed.
Mark and Jim, I wish you the best; but, honestly, I fear the worst. I only say this because I've seen it many times before — optimistic AI projections turning into bitter disappointment — time and time again. Nonetheless, if you manage to deliver half of the moon you promise, I will be the first to eat my words. (If only you could be more convincing at this stage of the game...)
Good luck,
-Phil
Notes:
1. There are two very different "contact" pages on the Machine Intelligence Technology website. The one accessible from the home page gives a street address; the one accessible from the blog does not.
2. As a one-man enterprise, I can sympathise with Dr. Gouge's resistance to hiring outside help for the more mundane tasks. Once you extend employment beyond the corporate officers, all kinds of regulations and taxes kick in, not the least of which are state industrial insurance and state unemployment insurance. If the officers are provided with other benefits, such as a retirement plan, they may, by law, have to be extended to all employees as well. These are major expenses that can cripple a nascient business. Frankly, and unfortunately, the regulatory deck is stacked against hiring outside help.
3. In searching for the "many" patents issued to James Gouge, I was able to find five: numbers 6373984 (2002), 6067371 (2000), 5040225 (1991), 5224175 (1993), and 5267328 (1993). Did I miss any?
You are continually explaining to us what Dr Jim's solution is NOT, but you can't seem to explain what it IS. All your attempts simply make it plain that you do not yourself understand it. This is fine - you have acknowledged as much in the past. But then we have Dr Jim's own attempts to explain his ideas - his videos - which appear to be equally unintelligible to most of the people who watch them.
Here's the situation as I see it, and then a suggestion for you:
Your continual references to mainframe computers appears to be confusing just about everybody. Perhaps you could tell us exactly what it is you believe is so unique about mainframes? It cannot be their unparalleled (pun intended!) ability to time-slice or context-switch - these are exactly what you say you DON'T need. It cannot be their parallel processing abilities - many mainframes did not have true parallel processing at all (or if they did it was often for fault tolerance, not for processing capability). Perhaps the paradigm you really mean is the supercomputer - where true parallel processing was built in from the start because it was necessary to solve the computationally intensive problems they were designed for. But then you say computational power is not what you are after (which is just as well because the Prop doesn't have much). But computational power is EXACTLY what both mainframes and supercomputers were built to offer - lots and lots of grunt, just aimed at two different problem domains. So we are naturally a little confused as to what it is you really mean. Please explain.
Then there's Dr Jims videos. Dr Jim gives a slightly unorthodox explanation about the functioning of neurons and synapses in the brain, and explains that the seat of learning is the synapse. Fine - most people can agree with him there. But then you claim that his solution is NOT based on neural nets. But on the other hand it IS based on a biological model. So it's biological, but not based on either neurons or their interconnections (which is all a neural net is). Now let's just disregard the apparent discrpency between what Dr Jim claims and what you claim - perhaps you could instead explain exactly what biological model you are referring to? Natually we all jumped to the conclusion that you mean the brain - but perhaps not. Dr Jim obviously believes the seat of LEARNING is in the brain - are you trying to say that the seat of COGNITION is somewhere else, or based on another biological model? If so, where? And what?
Answering the above points would be a good start - but that's not actually my suggestion. My suggestion is that you get Dr Jim onto these forums (as was previously promised) to answer some questions - without yourself trying to act as an intermediary. We promise to behave, and Dr Jim can even pick the topic of the day. Here are some possibilities:
- his software architecture (KISS OS, mainframe addressing etc)
- his hardware architecture (his memory board etc)
- his ideas about learning, cognition and machine intelligence
- why he believes the Propeller is the right solution for his particular needs
Some of these topics may be more appropriate for the Sandbox than this forum, but you choose the topic and the venue. I'm sure you'd have an enthusiastic audience.
Ross.
P.S. Has anyone in these forums actually purchased products from Mark or Dr Jim? If so, how about posting a review here? Why did you buy it, what are you using it for, and what do you think of the product so far?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Catalina - a FREE C compiler for the Propeller - see Catalina
I'm trying very hard to remain neutral until I see a demo
... hands slipping, limestone cracking, death looms on canyon floor [noparse]:)[/noparse] ....
Still ...
Apparently, the Robot magazine articles have been responsible for a surge in Propeller DIP40 sales.
Whether or not the memory boards have enjoyed proportionate sales is unknown.
If Dr Jim turns out to be a Hubert Farnsworth (whom in more than one Futurama episode
was credited with inventing the first anthropomorphic robot) and delivers, then that's great.
If not, at least the line of "Class Action Suit Lawyers" (sounds like GI-Joes) won't be aiming at Parallax.
BTW: Bill Henning is an excellent speaker.
Another forum participant who's demos are like sleeping pills could seriously benefit by offering Bill a consulting gig.
I think trying to build AI with Props is kinda like trying to build a nuclear reactor with Legos...
You might have some fun with it, but I think the prospects are very limited...
I think the nuclear reactor would be easier. Lego parts are water tight. If you made a chamber filled it with heavy water and made some tubes to run the uranium pellets through you would have a basic reactor
Even if you can get the parts I don't recommend you try. Would not be safe.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us and propmod_1x1 are in stock. Only $30. PCB available for $5
Want to make projects and have Gadget Gangster sell them for you? propmod-us_ps_sd and propmod-1x1 are now available for use in your Gadget Gangster Projects.
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
jazzed said...
Apparently, the Robot magazine articles have been responsible for a surge in Propeller DIP40 sales.
Jazzed,
I really doubt this - or perhaps I merely hope it isn't true
But if it is true, then at least it's good news for Parallax - and perhaps some of those disillusioned buyers will find these forums, and thereby discover all the wonderful things they can do with their purchases.
Ross.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Catalina - a FREE C compiler for the Propeller - see Catalina
Comments
Dr. Jim's code is very tight. It needs more than 32k, but can fit on 4-16 MB easily, 2 if he squeezes very tight and holds his breath while using a shoe-horn. He is using the Propeller for the cogs. Each cog is fast enough to do what he needs it to do. It is all about parallel design. He comes from the mainframe world and is building a sort of "mainframe" architecture for this project.
I hope this clarifies a little.
Mark
Very boring clips. I didn't hear anything that was grossly wrong, but the general impression was that of a lot of not very well connected, but well known information with nothing of any depth about the supposed subject of the talk. There was nothing I heard that wasn't written about in books and magazines years and years ago, and discussed much better.
I've worked in the area of "holographic" machine memory in the past and the same kind of pitches were used then. Technology is much better now and we were working with much slower and smaller computer systems than the expanded Propeller system you're talking about. Unfortunately, real machine intelligence is much much more complicated than undifferentiated neuronal networks and we're only just beginning to understand some of the deep organizational structure involved in the brain.
You may have some great ideas, but your skills at presenting them need work. Some successful bright people with great ideas find others with the technical presentation skills and provide them with the information needed to do their job. Right now unfortunately you and Mark sound like snake oil salesmen and these video clips don't help. Saying "it ain't so" and "just wait, you'll see" doesn't make it different. It would be a terrible waste to have good ideas and just a lousy presentation.
We have a following that thinks we are on to something, and it is for those people we will continue to do our best to convey the technology as best we can within our budget.
Time will tell what the outcome will be.
Mark
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us and propmod_1x1 are in stock. Only $30. PCB available for $5
Want to make projects and have Gadget Gangster sell them for you? propmod-us_ps_sd and propmod-1x1 are now available for use in your Gadget Gangster Projects.
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
Our numbers are confidential, however, our following are buying without seeing the finished product. They are taking a risk, to be sure, but the great thing about it is, they have enough confidence in our approach to purchase our products so they can become familiar with them before the project is finalized. These are our early adopters, and they will have a head start when the wider public starts accepting our approach.
The wider public will come on board when they see machine intelligence on the propeller, first with acceptance of voice recognition as we prove out the concept and show video clips of it on our blog, then as we add binocular vision and response, tactile input and response, ultrasound object detection and response, motor functions, etc. When this is all working, we will have a large audience indeed.
And we are on the cusp of all of this. You are entitled to your opinion, but mine is that we will succeed.
Thanks,
Mark
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us and propmod_1x1 are in stock. Only $30. PCB available for $5
Want to make projects and have Gadget Gangster sell them for you? propmod-us_ps_sd and propmod-1x1 are now available for use in your Gadget Gangster Projects.
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
Lot's of people work on a shoe string budget. I'm sure that you have a university or college near by with lots of students willing to work for cheap while still having the technical knowledge to make a good presentation.
As a side note, I was going to suggest a couple of places around where you are that could supply some students. Unfortunately, your website doesn't list any street address or even a telephone. I have to ask: what reputable company only allows communication via the web? A PO box only costs about $60 a year, and a telephone number isn't much more.
Edit: Technically, I'm certain that any business (at least in California) has to have a registered address and possibly a phone number on file at a public registry. So, therefore, you have an address that is floating around out there somewhere. Why not share it? I do hope that you incorporated to prevent a lawsuit from taking personal property.
Post Edited (SRLM) : 7/31/2009 7:00:00 PM GMT
Those are some good suggestions. But Dr. Jim has his own ideas how he wants to proceed.
If you look at our Contact Us page you will see our physical address.
Thanks,
Mark
There is certainly nothing wrong with coming home from your real job to do your own designs, take web orders, and put stuff together in your garage for sale by mail-order; several people here are very openly doing that and sometimes they offer real bargains, as when they've done the SMT for us or sourced components that are a lot easier to get in quantity. And let's face it, just about everything that exists so far for the Propeller was built first in someone's garage.
But there is something wrong with pretending to be something you're not. Designing a memory expansion is not special; there's been a lot of discussion here about how to do it. Popping in to claim that your memory expansion is the bestest thing ever when you aren't even aware of those discussions impresses no one. Designing an operating system is not special; there are half a dozen similar projects in the works, some with significant progress. Again, the relative merits have been discussed extensively and all of them will have a place, because they are aimed at different goals. Designing hardware I/O controls is not special; hell that's Parallax's whole business model and half of the contents of the obex.
So we have items A, B, and C, are reasonable enough to assume they have emerged from your garage of businessdom in relatively non-vapor form. Then we have item D, wherein you claim that items A, B, and C will allow you, with the aid of a few eight dollar chips, to solve a problem that has continuously beaten very smart people who are paid well to spend all of their time going at it with supercomputers and hundred thousand dollar robots. And that furthermore the imminence of item D implies that items A, B, and C must be highly desireable and worth considerably more money than similar products offered by other garage dwellers.
Furthermore, when pressed about item D, you reveal that you are ignorant of what is actually known about the biological solutions that exist, you're ignorant of current research in the field, and you're even ignorant of a fairly prominent philosophical movement which is OBSESSED with item D that has generated several bestselling books and thousands of blog posts. All right, not a lot of people outside of the Singularity community have heard of Eliezer Yudkowsky, but surely you've heard of Ray Kurzweil? And the DARPA challenge? You do realize the people who won the DARPA million did it by doing almost exactly what you claim you will be doing?
I am someone who has thought about AI a lot. I have been convinced since I was a child -- in the 1970's -- that strong AI is possible and will eventually be created. I also believe it will be self-programming through experience and much of the reason it hasn't been done yet is that people in the field are too focused on specific results instead of trying to get results to emerge from chaotic interaction.
But I do not believe you are going to get a machine that learns English on its own and asks sensible questions from a platform with a few propellers and 16 MB of RAM. I don't believe it because if it were that easy, it would have been done already on a more expensive and powerful platform. The prop is very clever, but we're not here because it's the fastest thing out there; a lot of DSP chips could run circles around it, 8 cogs or no, and a bottom of the line desktop PC also has more power even with the context switching overhead. And that's not even to mention things like GPU cards and the Cell processor.
So you have a memory expansion and some I/O routines, very nice, they might be useful in a lot of ways. But please forgive us if we are not impressed when you claim that they are our chance to get in on the ground floor for owning our own Lieutenant Data. Or even Skynet.
Having said the above, the Japanese are well ahead of you in this field, they already have robots (machines) that can do just about everything that has been talked about here, and soon enough they will be marketing here. So, time is not on your side, you may have an excellent product, but what good is that if they beat you on your own turf.
You just have to stick with it, most of these skeptics are not going to purchase your product anyway, so do not fall for there baiting schemes.
Now, that the SX chip is EOL I will have to start thinking more about the Prop chip. Since you guys are doing the voice, I may have to go for the replication of striated muscle function.
Ray
That's cool! We need striated musculature in our bots!
Like I said, what Dr. Gouge is attempting may have been done before a different way. That's ok by us. We are attempting to do it our own way. This should not be the only way, but we feel it is a good way, so we are going for it. I know it is controversial because almost no one has figured out how to do it well. We are taking our stab at it. Mabe one of you will have an even better idea than what we develop. That's ok too. We would love to see it. But we are proceeding with our project just the same.
I have seen some of the things the Japanese have done. Some of their work is quite impressive, but still appears to be programmed. I saw a model that cost $50,000 for a robot and was nothing more than some fancy hardware and software, and a remote mic where a person sitting across the room could hear and respond to people's questions. That, I believe we can all agree, is fraud.
Mark
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Toys are microcontroled.
Robots are microcontroled.
I am microcontroled.
If it's not Parallax then don't even bother. :-)
·
Mini-Din/PS2 connectors are for sale! 5 for $1! PM me if you wish to make an order.
Cheap·shipping unless specified!··········150 left!!··
Post Edited (microcontrolled) : 7/31/2009 9:58:43 PM GMT
If you don't want the complexity of·a multi-tasking operating system·you buy a basic PC, reformat the hard drive, and install FreeDOS.· You now have a single tasking system that will run rings around a Propeller and also supports lots of·fancy PC-based I/O devices such as hard disks and cameras.·The Propeller is not more powerful than a PC.· I have programmed both, and it's just true.··It's also not harder to·program if you know what you're doing; you can build your own multitasking state machine engine·for a lot less effort than the complexity of rigging up megabytes of RAM to a propeller.· The PC will use a lot more power but it doesn't even cost that much more when you factor in the·available I/O buses and memory.· We are not here·in Propellerland·because the Prop is more powerful than a PC, we are here because the Prop can generate video in software (which in turn has more to do with deterministic timing than superfast timing), its I/O pins are general purpose and can·source/sink 40 mA, and it runs on AA batteries.
Let me be perfectly clear on this; I have done this kind of programming.· For awhile I had my eye on the DARPA challenge.· (That lasted until I realized that if I succeeded, building SkyNet and giving it to the very people who were most likely to turn it into SkyNet was probably a bad idea.)· I agree in principle with many of Dr. Jim's stated ideas.· But I am not here because I think the Propeller is the key to building a Seed AI.· I'm here because the Propeller can really reduce the parts count and design complexity of embedded devices.· It occupies an otherwise very empty corner of the price / performance map in that respect.· It is not, however, supercomputer material, and I strongly suspect Chip Gracey would agree with me on that.·
There are other things out there that are low-cost candidates for this project, such as GPU cards and the Cell processor in the Sony PS3, both of which can run circles around an entire farm of Propellers.· With the Prop and the memory board I will grant that you might get something that can find its electrical socket when the power runs low.· But the very idea that it will learn natural languages to communicate with humans is not just laughable, it's stupid.
The reason I chimed in on this is that I detect an unfortunate pattern here.· The business operated out of a garage pretending to be bigger than it is is no rarity and not much source of shame, though I admit I have more respect for people who are more up-front about their real capabilities.· It's the hyperbolic claims, the selling things sight unseen based on these hyperbolic claims and perhaps a cult of personality, that worry me.· This reminds me very much of a scam that was pulled in the late 1970's when semiconductor RAM was hyper-expensive, lots of computers still used core RAM that was created by manually threading wires through teeny ferrite donuts to create a bit,·and there was a technology called "bubble memory" that was supposed to be the Next Big Thing.
Several companies jumped on the bubble memory bandwagon, attracting investors by promising that their bubble-memory based tech would create exactly the same things Dr. Jim is promising -- computers that would talk with you, play chess and beat you (that hadn't been done for real yet), even control android robots.· Bear in mind that in that day a megabyte of RAM was the kind of extravagance only a government or multinational corporation could afford, so this convinced a lot of people.· There were VHS videos and public demonstrations where a "bubble memory controlled" robot would play chess with audience members.
Turned out it was all a scam; the promising demos, including the chess-playing android robot,·had been faked (it was remote controlled by an actual human chess master), and the investment money disappeared along with the founders when the lie got too hard to hide.
I hope that Dr. Jim is just naive about the possibilities for his self-educating network; it's an approach lots of other people are working on with far more capable hardware though.· But the possibility that he has plugged "parallax propeller" into an old, old script in place of "bubble memory" would be a much uglier scenario.
First, we have to get one thing straight. We do not require massive computational power. We require a massively parallel system, which is why we need the cogs, which is why we need multiple Prop Proto boards. Each cog is fast enough to do each function necessary, but you have to overcome this idea that massive computational power is needed. That is NOT the case with our approach.
Dr. Gouge wrote the telemetry for the LEM (lunar lander) during the Apollo missions. They had to page up 4k at a time due to the restrictions of memory at the time. It was a mainframe world back then. As a result, he writes very tight code and thinks in terms of mainframe architectures.
Since he is able to create his own platform his own way, he has plenty of power in each cog to do what he needs to do. He simply needs many cogs to do it. Again, get away from PC architectures and look more to mainframe technologies and you will understand our approach a little better.
We are not doing any kind of remote brain using RF or microwave, etc. because it is far too slow to do what we need it to do. We have chosen static expanded RAM to become the neuronal structures because they are very fast and will be able to react quick enough for Dr. Gouge's machine intelligence algorithms. And as Dr. Gouge explained in his videos, expanded RAM was necessary, or this project would not be able to get off the ground.
If you wanted to use a PC chip approach, you would need 32 to 64 computers to work out the same approach. The cost would be prohibitive for the average hobbyist and far too bulky for our purposes. We need to be mobile. As it is, you will end up with a "cube" of processors and memory that will need to be installed onto the robot.
You are absolutely correct about Propeller pricing. This is a major factor why we chose it and why we can use multiples of the Prop Proto board for the COGs we need for our project. Four or eight boards is within the price range of anyone desiring to do this project on a kitchen table.
But again, we are not building a Supercomputer. The architecture is more like a mainframe. Supercomputers need to be super fast. We do not need that. What we need is massively parallel. So again, think mainframe, not supercomputer.
The other hardware that you mention, such as GPUs or PS3s would also be prohibitive cost wise for the same reason, you need multiples of these to do the trick, and you do not want to mess with multithreading etc., way too much overhead. COGs work perfectly and at the right price for our intended architecture. Believe me, we thought this through when we chose the Propeller.
About learning natural languages...just wait and see. You may be surprised.
While we may be working out of the garage (spare room that we turned into a lab), we are not liars or scam artists. You will believe what you want to believe and that's fine. But keep an open mind.
From many of these posts, I still see too many people thinking the wrong way in a PC world where faster is better. We don't need super fast processors. Just fast enough processors and a lot of them. Start thinking about mainframe architectures and you will be thinking much closer along the lines of what we are doing.
Great thoughts, though. Keep them coming.
Mark
This is where so many people go wrong. Throwing a PC at the problem is not always the best way to do something. Back when I was younger, we didn't have all of these powerful microcontrollers to run things. We had to figure out clever ways to do things. Look at the robots from hundreds of years ago - they didn't even have electricity, yet they managed!
A PC with a 1TB hard drive and 8G of RAM is a sure way to guarantee that the bloat will eat up your project. And using MS-DOS or (worse), Windows? Don't even get me started! The power requirements of a PC are horrendous. All of that bloat requires more battery power, which requires a more powerful robot to carry it, which requires a bigger battery to move that heavier robot. Going with a clean slate, and starting with a processor that was designed for parallel processing, instead of having it tacked on, along with tools that are designed to support it, and not to sell bloated consumer code to dullards, seems like the more sensible way to go.
edit: I started this post before the above post was posted. Looks like we're thinking along the same lines. Think Different, people! Get out of your mental ruts, and the world will open up for you!
Post Edited (Marka32) : 8/1/2009 1:14:10 AM GMT
Please stop trying to impress me with references to mainframe computers. When in college I was taught FORTRAN on a DEC SYSTEM 10, which if I'm not mistaken is a mainframe computer. I have also done real work on 70's-era 8 bit PC's, and more recently on similarly powered equipment like the 8051. I know all about tight code and doing it with integer math in assembly language. I am 45 years old and I have all those T-shirts in my closet.
Massively parallel systems are about computational power. Saying they are not about that is like saying Beyonce is not about sex. The Prop has 8 cogs because it would have to run at 640 MHz to give equivalent computational power without the parallelism, and it's a very conservative design for these days that keeps the frequencies out of the UHF range. A similar non-multiprocessor solution running at 640 MHz should be able to do what the Prop does at 80MHz. In reality, there are other factors, such as context switching and the fact that cogs are really, really fast because of their 2K cog memory limit which mean the other systme would have to be a bit faster to emulate a Prop, but when common PC's are running 2 GHz processors and megabyte on-chip RAM cache, it's very silly to think a propeller can do something non timing-dependent that a PC can't.
You do not need 32 to 64 PC's to do what one propeller can do; you need one. The one thing the Prop can do with absolute constancy the PC can't is generate timing and waveforms. The PC will be as fast in processing but not as reliable on a microsecond level. This is of absolutely no consequence in the kind of software you are talking about. It's very important if you want to generate video or RS232 signals in software, but irrelevant in neural network simulation.
I'd be interested in an actual explanation (with actual numbers and examples) of why RF remoting isn't practical, when RF cameras delivering realtime video over wifi are under $200. I am pretty familiar with the entire range of sensors you might be thinking of installing on your android, and especially considering that your financing seems to be on the same level as mine, I can't imagine any combination you could afford that would saturate a wifi link.
What you don't seem to understand is that the propeller is, by modern CPU standards, SLOW. Cogs do not change this. CMOS gates use energy when they change state, so 8 cogs at 80 MHZ will use the same energy as one similar single-cog CPU at 640 MHz. That is still slow by modern standards, and there are many DSP chips that blow smoke rings around the Prop while being designed for things like processing video.
As for your product learning natural human languages -- I will wait, but I dont' think I will be surprised. I will be the first to salute and offer apologies if I'm proven wrong.
Not to hijack the thread, but this is a significant bugbear of mine. IP cameras over WIFI are *not* real time. There is significant latency which poses a number of challenges to the designers of CCTV systems with IP cameras. They are also unreliable and can be easily interfered with. Both less than desirable qualities (slow and unreliable) when dealing with CCTV.
I'd sure as heck hate my Robot to be slow and laggy when it's trying to pickup a crystal scotch glass.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
lt's not particularly silly, is it?
I respectfully disagree with almost everything you had to say.
Anything I could say, you will disregard, and likewise, most of what you had to say is not applicable to our machine intelligence technology, and many of your facts are wrong, so we are at an impass, with due respect. For example, DEC never made a mainframe. They made MINI computers and MIDI icomputers, but never mainframes. A DEC PDP-8 MINI computer had a fortran compiler as well as cobol, et. al. That does not make a PDP-8 a mainframe. PDP-10s and 11s were mid-range computational platforms, or MIDI computers. If you want to talk about mainframes, there were only a few. Mostly, Univac, IBM, and Control Data Corp. These were the major mainframe suppliers of the time.
Our technology is not based on any neural-net paradigm, so the computational power needed for those is not required for our architecture. We are more interested in the parallelism, not having to do context switching. We don't need time-slicing, but rather true multiprocessing, which the Propeller provides nicely.
If you are looking for an example, look to the biological model, not previously tried technologies based upon neural nets, fuzzy logic, etc. Those have all been tried for decades with limited/questionable results. As an example, there is one question to answer. Where is the truely intelligent robot? It has not yet been made.
Our technology promises to provide a major step in the direction of autonomous, intelligent machines and provides an excellent, low-cost platform for future research. Isn't that what it's all about?
Respectfully,
Mark
I can see space being an issue. A small pc is 4"x5" my brainboard is 1"x7l but since you are using dev boards and dip ram I assume that is not a care of yours.
Cost my brain board is $200 pc is about the same.
I hope it works. Your system is to big but I would be willing to work with you or buy a license to make a smaller version for myself
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us and propmod_1x1 are in stock. Only $30. PCB available for $5
Want to make projects and have Gadget Gangster sell them for you? propmod-us_ps_sd and propmod-1x1 are now available for use in your Gadget Gangster Projects.
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
We are using the Prop Proto USB boards.
Dr. Jim knows the details of our architecture way better than I do, but I can see the advantage of having 8 propellers on one board (forgive me if I am stating this wrong). Perhaps it would be good to collaborate with Jim on merging your multi-prop board with our expanded memory and that way we would have 64 COGs at once on a much smaller physical footprint.
I know Dr. Jim wants to use the serial mouse to talk to multiple Prop Proto boards. He has reasons for using multiple boards, for use of additional I/O pins (which you could probably access via your 8 chip system just as well) but also for USB devices for thumb drives, etc. Even if we could not use it for all the boards we might need, it would make a superb "Master" board if we could work it into our OS. It's still probably worth asking him about it though. You might shoot him an email at support@machineinteltech.com. I'll make sure he gets it.
Also, if you know where we could get parts to reduce the cost of our ERAM boards, we would be interested in lowering prices so more people could get involved.
Thanks,
Mark
Post Edited (Mike Green) : 8/1/2009 3:39:23 AM GMT
apparently you never heard of a DEC VAX; a commonly
used mainframe. i know because i programmed on one
twenty years ago.
you and your friend jim seem to ignore common sense.
but what really bothers me (and apparently a lot people
on this forum) is you write a lot of words without saying
anything at all. i find hard to believe anything you say,
especially that you sold ANY of your wildly overpriced
memory boards.
tired of all this
blake
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"Can't sleep, clown will eat me."
Bart Simpson
Mark and Jim, I wish you the best; but, honestly, I fear the worst. I only say this because I've seen it many times before — optimistic AI projections turning into bitter disappointment — time and time again. Nonetheless, if you manage to deliver half of the moon you promise, I will be the first to eat my words. (If only you could be more convincing at this stage of the game...)
Good luck,
-Phil
Notes:
1. There are two very different "contact" pages on the Machine Intelligence Technology website. The one accessible from the home page gives a street address; the one accessible from the blog does not.
2. As a one-man enterprise, I can sympathise with Dr. Gouge's resistance to hiring outside help for the more mundane tasks. Once you extend employment beyond the corporate officers, all kinds of regulations and taxes kick in, not the least of which are state industrial insurance and state unemployment insurance. If the officers are provided with other benefits, such as a retirement plan, they may, by law, have to be extended to all employees as well. These are major expenses that can cripple a nascient business. Frankly, and unfortunately, the regulatory deck is stacked against hiring outside help.
3. In searching for the "many" patents issued to James Gouge, I was able to find five: numbers 6373984 (2002), 6067371 (2000), 5040225 (1991), 5224175 (1993), and 5267328 (1993). Did I miss any?
"What we have here is a failure to communicate"
You are continually explaining to us what Dr Jim's solution is NOT, but you can't seem to explain what it IS. All your attempts simply make it plain that you do not yourself understand it. This is fine - you have acknowledged as much in the past. But then we have Dr Jim's own attempts to explain his ideas - his videos - which appear to be equally unintelligible to most of the people who watch them.
Here's the situation as I see it, and then a suggestion for you:
Your continual references to mainframe computers appears to be confusing just about everybody. Perhaps you could tell us exactly what it is you believe is so unique about mainframes? It cannot be their unparalleled (pun intended!) ability to time-slice or context-switch - these are exactly what you say you DON'T need. It cannot be their parallel processing abilities - many mainframes did not have true parallel processing at all (or if they did it was often for fault tolerance, not for processing capability). Perhaps the paradigm you really mean is the supercomputer - where true parallel processing was built in from the start because it was necessary to solve the computationally intensive problems they were designed for. But then you say computational power is not what you are after (which is just as well because the Prop doesn't have much). But computational power is EXACTLY what both mainframes and supercomputers were built to offer - lots and lots of grunt, just aimed at two different problem domains. So we are naturally a little confused as to what it is you really mean. Please explain.
Then there's Dr Jims videos. Dr Jim gives a slightly unorthodox explanation about the functioning of neurons and synapses in the brain, and explains that the seat of learning is the synapse. Fine - most people can agree with him there. But then you claim that his solution is NOT based on neural nets. But on the other hand it IS based on a biological model. So it's biological, but not based on either neurons or their interconnections (which is all a neural net is). Now let's just disregard the apparent discrpency between what Dr Jim claims and what you claim - perhaps you could instead explain exactly what biological model you are referring to? Natually we all jumped to the conclusion that you mean the brain - but perhaps not. Dr Jim obviously believes the seat of LEARNING is in the brain - are you trying to say that the seat of COGNITION is somewhere else, or based on another biological model? If so, where? And what?
Answering the above points would be a good start - but that's not actually my suggestion. My suggestion is that you get Dr Jim onto these forums (as was previously promised) to answer some questions - without yourself trying to act as an intermediary. We promise to behave, and Dr Jim can even pick the topic of the day. Here are some possibilities:
- his software architecture (KISS OS, mainframe addressing etc)
- his hardware architecture (his memory board etc)
- his ideas about learning, cognition and machine intelligence
- why he believes the Propeller is the right solution for his particular needs
Some of these topics may be more appropriate for the Sandbox than this forum, but you choose the topic and the venue. I'm sure you'd have an enthusiastic audience.
Ross.
P.S. Has anyone in these forums actually purchased products from Mark or Dr Jim? If so, how about posting a review here? Why did you buy it, what are you using it for, and what do you think of the product so far?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Catalina - a FREE C compiler for the Propeller - see Catalina
... hands slipping, limestone cracking, death looms on canyon floor [noparse]:)[/noparse] ....
Still ...
Apparently, the Robot magazine articles have been responsible for a surge in Propeller DIP40 sales.
Whether or not the memory boards have enjoyed proportionate sales is unknown.
If Dr Jim turns out to be a Hubert Farnsworth (whom in more than one Futurama episode
was credited with inventing the first anthropomorphic robot) and delivers, then that's great.
If not, at least the line of "Class Action Suit Lawyers" (sounds like GI-Joes) won't be aiming at Parallax.
BTW: Bill Henning is an excellent speaker.
Another forum participant who's demos are like sleeping pills could seriously benefit by offering Bill a consulting gig.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
--Steve
Propalyzer: Propeller PC Logic Analyzer
http://forums.parallax.com/showthread.php?p=788230
Post Edited (jazzed) : 8/1/2009 4:43:17 AM GMT
You might have some fun with it, but I think the prospects are very limited...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
My Prop Info&Apps: ·http://www.rayslogic.com/propeller/propeller.htm
Even if you can get the parts I don't recommend you try. Would not be safe.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
propmod_us and propmod_1x1 are in stock. Only $30. PCB available for $5
Want to make projects and have Gadget Gangster sell them for you? propmod-us_ps_sd and propmod-1x1 are now available for use in your Gadget Gangster Projects.
Need to upload large images or movies for use in the forum. you can do so at uploader.propmodule.com for free.
Post Edited (mctrivia) : 8/1/2009 7:07:39 AM GMT
I really doubt this - or perhaps I merely hope it isn't true
But if it is true, then at least it's good news for Parallax - and perhaps some of those disillusioned buyers will find these forums, and thereby discover all the wonderful things they can do with their purchases.
Ross.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Catalina - a FREE C compiler for the Propeller - see Catalina