Hey folks,

I intruded on another conversation, and thought I would move it over here instead. I was interested in the posts/string on wide data 65xx, because I am part of the "new generation" of computer fanatics; the generation of users, the generation that wants more functionality without actually understanding how to get it. Luckily, I am a scientist, so I treat everything with twice as much skepticism as it deserves! I am a fan of optical computing! when will the revolution arrive? I am not a fan of quantum computing, as it has never made sense to me. It will always be mysterious to me.

I am quite interested in philosophy, computer architectures, natural language, computer language, syntax, semantics, and a hundred other philosophical topics of interest to the computer programmer, designer, or user. I am a huge proponent of STRONG AI, and, in particular, I am a big fan of one theory for obtaining such results, using the mere machines presently at our disposal (humbly enough, I called it "Meyer's Law".).

So, while I appreciated the FPGA discussion about wide-byte systems, I definitely prefer to talk about future processors. 3-D chips have been suggested for probably 20 or 30 years now; but the heat must be a nightmare. Perhaps, some have designed around the problem? The transistors will shrink, for a little while longer, but soon the poor electrons will leak between the atoms. And then where will Moore's Law be?

My guess is some kind of 3-D architecture and/or massively parallel computing. With regard to the data-width the processor should operate on, maybe flip chip could package more static RAM into the same small space? Or maybe Multi-chip modules could do this? I don't know. I don't have the background or experience in hardware design, or even a consumer level knowledge of what is presently out there, on the market, in terms of available RAM (i.e. fast RAM; on the order of 5 to 20 ns read time).

3-D hologram computing seems promising, too. In fact, I think--in my opinion--some of the first analog computers were the lenses developed for the Daguerreotype, circa 1850-1860. The first photographic lenses were just re-purposed telescope objectives, and it took a contest and a man named Petzval to make a lens suitable for long exposures and a narrow field; i..e portraiture. But lenses are, most certainly, adept at doing a "near infinite" number of trigonometric calculations, in "real time" at the "speed of light". How does one compete with this? How does one harness this understanding and merge/meld this capability into the concept of the "general purpose, re-programmable thinking machine" that we know and love; i.e. the "computer".

I understand, I am talking "fuzzy talk" right now, but I could go much further and into a fair amount of detail. SO I thought I would post and throw this question out there. I was a little bit sarcastic when I suggested we wire up an iphone with a wide-byte 6502; but I might have been a little bit serious.

Here is a free idea (so long as you always tell people where you got it!). I can't afford the cheap machinery to get this dream tested, but maybe you can. How about an optical 6502? Before I get to some notes on the practical implementation, how about a little bit of theory. Charles Sanders Pierce, one of the founders of the philosophical movement of pragmatism, the only philosophical school native to the United States, circa 1933, proved that all combinational logic circuits could be constructed by NOR gates (and subsequently it hasd been proven, likewise for NAND; My Wakerly Hardware Textbook mentions that 3 are necessary; Inverters, AND, and OR gates.). So, leaving sequential circuits (flip flops and latches) aside for the moment, it might be reasonable to assume that an optical circuit capable of keeping signal fidelity over millions of repititions (or maybe just tens of thousands, before the signal is stored and/or refreshed and/or "re-constuituted") could be constructed in any system that can make one of these crucial logic gates.

So, now the nuts and bolts (I will describe them; you can feel free to build it without my extraction of royalty payments!). I have read a hundred article that claim to crack it (i.e. optical computing); mostly trade magazine stuff. I have never read one that tried this method. Try this one, for free; (1) Fiber Optic "wires", "traces", "signal path", (2) two separate laser wavelengths (i.e. colors) carry the signal, (3) a series of dichroic mirror filters (some reaching reflectivity of 95 to 99 percent) comprising the logic gate apparatus, (4) a Blue Ray DVD "CD-jukebox", (5) many lasers, for massively parallel read from the BR-DVD, and (5) massively parallel computer architecture (a million-headed-6502, if that is your desire!). It is just a thought. You could start small, and work up to the million headed-hydra-6502!

So, just two problems (or 3?) to crack, here. (1) exact implementation of color filters to produce the digital logic (shouldn't be a problem for a clever guy/gal!). (2) How to store light (i.e. feedback circuit, at least, as in the case of the flip flop, latch, etc.). (3) exact implementation, as regards signal fidelity (i.e. light is funny stuff, and likes to leak out, absorb, and not stay where an engineer puts it! This, I am sure, can be overcome by crafty engineers!). And, I guess, one last concern; can it be scaled as small as present and/or future transistors? (does it need to be? a massively parallel computer MIGHT stilll add more and better functionality, even at the price of space, time/operating speed-frequency, or power/heat-dissipation).

Now, why mention Bender and the Terminator (both, it is rumored, ran on a 6502)? Well science fiction and science fact often overlap, and it is my intention to post to this string if any foolishly clever ideas occur in the future. I hope you feel free to do the same. IN the mean time, if you ever get that optical multi-headed-hydra-6502, wide-byte processor, running, do let me know; my terminator might need one.

Scratch that "short exposures and narrow field" ... hard to hold a smile that long. In the early days the film wasn't too fast, either. Slow lens and slow film necessitated a better lens n(and better film) for moving subjects, i.e. people!

Regarding:

"(2) How to store light (i.e. feedback circuit, at least, as in the case of the flip flop, latch, etc.)."

CCD was originally though up as a analog storage medium (or was it digital storage?). Circular Charge coupled devices could (not light sensitive) acted as delay lines, and transmitted information, stored values, serially, at a slower speed so that the other parts of the computer could keep working. Then the data was delivered when it was needed.

#2 could easily be solved by something like that (weren't there, in the old days, mercury delay lines? Seems rather severe a solution? Oh well, I guess it made as much sense as Faraday and his "mercury motor"; the brush commutator would wait for Sturgeon/Davenport/Jedlik).

Also, on the subject of Analog computers (i.e.2 posts above; lenses and analog trigonometry calculations via Ptolemy's-rule-of-thumb/Snell's-Law), Albert Michelson, despite being known better for other more cosmically important things, could solve fourier analysis and fourier additions with his harmonic analyzer (like Lord Kelvin's tide predictor, before him. And, I suppose like John Harrison, the clockmaker/marine-chronometer, before him!).

There's been talk of it for decades; but after having just checked the Wikipedia article on it, I won't hold my breath. It's not lookin' good.


See my links at http://wilsonminesco.com/links.html#philosophy . See also my post about the book "Thinking Forth" which is really more about programming philosophy.


There's plenty on this site and on these forums. You don't have to get into superhigh-powered processors to get into these. On the contrary, my view is that the more power is available, the more it gets wasted, justifying extremely inefficient, sloppy code.


The newest interest of Chuck Moore (the inventor of Forth who turns 77 this year) seems to be http://www.greenarraychips.com/ where he has 144 stack processors on a single IC, for massively parallel embedded applications. Each of the processors on the IC runs Forth instructions in as little as 1.4ns, putting equivalent peak performance at 100GIPS, ie, a hundred billion instructions per second.


I offer a hobbyist-friendly 4Mx8 5V 10ns SRAM module, intended especially for use with the 65816. Data sheet at http://wilsonminesco.com/WM-1_4Mx8SRAMm ... -15-13.pdf , and forum discussion at viewtopic.php?t=1908 /

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

That seems awesome. I might be a customer for you RAM!

I am still holding out for the optical computer; I don't care if it fills a room, at first. I am especially curious if the energy dissipation can be brought down, allowing future diminution of the structures. That is why I decided some sort of color scheme was warranted; since dichroic mirrors seemed very efficient (99% SEEMS like a good number, until you calculate how many reflections reduce it to nearly 0% of the original signal!). Usually wikipedia is my first stop for answering such questions, but I live in the real world, and my lasers and the few dichroic mirrors I have are just begging me to try this out!

As for frugal code, I don't even code yet, and I am already in support of it!

I will definitely check the link to massively parallel IC. GIPS seems a lot? I don't remember what is a big number in MIPS or FIPS, or in FLOPS? I'll take your word for it.

The frugal-est of code, might, in the end, prove to be the algorithms running through the human mind. 6,000 (approx.) human languages on earth, and all can be--roughly--described with just six possible structures (four , in practice, as there seems to be some native/genetic and/or cultural/historical bias against two possibilities. Could just be an accident of history? Probably that!). Linguistic Typology categorizes--roughly-- human languages as being SVO, SOV, OVS, OSV, VOS, VSO, for their permutation of the basic word order in the language; Subject-Verb-Object. This will be familiar to you; much like decoding memory. (A must read, if you are into philosophy, is Aristotle's "De Cognita et Reminiscienta"; but try it in English, the Latin/Greek will be ten times as impenetrable! Tell me he isn't describing RAM and ROM! OK, maybe I exaggerate, but he is pretty darn close to describing the basic functioning of any "remembering system".)

So Meyer's Law, which I so humbly alluded to, just calculates the maximum number of simple sentences that any artificial or natural intelligence can construct, given a number of known nouns and verbs. (I say "simple sentences" because the mere addition of words like "a" and "the", not to mention the numerous prepositions, expands this number by millions or billions, I am sure. However, prepositions and "a/the" can be "gleaned from context", a linguistic device the human animal is quite adept at utilizing.). I have a good study on the number of words children say in a day (studies conducted in the 1970s, when cheap tape recorders, and tape, became affordable) and it should be mentioned that a derivative could be utilized--rate of acquisition of new words, and/or rate of utilization of known words--extending Meyer's Law in several other directions; many of them fruitful to applications towards Strong AI.

I should mention Noam Chomsky, Daniel Dennett, Stephen Jay Gould, and Richard Dawkins as sources of inspiration for the above notions. Also, I think it bears mentioning that the Jaina philosophers, in 600?B./C. were working out the combination and permutation formulae, and its first application was to the number of tastes one could cook into a food dish! I suppose, when you come right down to it, that's what we are doing when we put all our words together in the right order; most of us are looking for a new and interesting dish! Was it Pascal or Fermat or Poisson who brought combinations and permutations to the western mathematical tradition? Or do I remember somebody earlier? Maybe Fibbonacci? I know he brought us "numbers' , numerals as we now know them! Awww heck, I forget!

Meyer's Law => VN^2=S ; Quite a simple concept, for so powerful a thing as speech!

It seems like a very dumb "AI" could generate all possible sentences given a fixed vocabulary, and game the metric you've presented. Consider if ELIZA added randomness to its responses, weighting higher for word combinations that haven't been used recently.

3d stacked RAM of course is hitting the market, but that's a completely separate process with its own more stable heat characteristics, as far as I understand. As far as 3d processors go, I'm not sure we're going to reach that soon, and I don't think it's necessarily a focus. A giant limitation is memory bandwidth. We have incredibly fast ALUs and FPUs, but it's juggling all the pipes carrying the data that's seeming to becoming a very fundamental conceptually limiting factor.

My favorite architecture is the Venray stuff, but they want to license the IP, not actually build the processors themselves, and haven't shown any traction in doing that. Basically, they created CPU tech on DRAM process, which seems like an ideal solution to me: http://www.venraytechnology.com/

Beyond that, the Mill CPU is an interesting rethink of CPU tech in general. It's basically a mass of ALUs with wide instructions telling which output should go to which input on each cycle. It also has a very unique programming model built on top of that. http://millcomputing.com/

_________________
WFDis Interactive 6502 Disassembler
AcheronVM: A Reconfigurable 16-bit Virtual CPU for the 6502 Microprocessor

Here's nice presentation about modern IC fabrication. https://www.youtube.com/watch?v=NGFhc8R_uO4

Seeing the immense scale of operation just to make 2D chips, I don't think the same technology will be feasible for full 3D.

Forgive my cursory response today; I will follow your links/references (@Arlet and White Flame) as I am woefully out of step with modern super-computing; too much happens too fast for me to follow it all!

I wanted to post these photos yesterday, but couldn't find them. Its the kind of project that only a biologist could try ("After dissecting the animal, little Timmy threw the lungs, and heart, and stomach back into the body cavity and sewed it up; but ultimately, he found himself standing there, befuddled, when it became clear that it would never move again"! This, in summation, is my attitude towards hardware design! Perhaps my attitude is about to change!).

I'll post three photos, but I think they are big files; I don't want to clog your server and/or data-farm(I don't know what sort of capacity your organization has?!) so i posted the link to my FB page.

https://www.facebook.com/randall.meyer. ... 604&type=3

Apparently I on't know how to post photos? i thought i knew? I will attach as files; maybe that is how I did it before?

So, two are my project (circa 2007-2010) and one is a placstci mount of a Multichip module from IBM (also photo-ed from the same period).


More later; Cheers,

(@White Flame)

"It seems like a very dumb "AI" could generate all possible sentences given a fixed vocabulary, and game the metric you've presented. Consider if ELIZA added randomness to its responses, weighting higher for word combinations that haven't been used recently."

Well, yes, of course. I never said that intelligence was anything special, just rare. I suppose it is special, for other reasons, though? N'est-ce pas? Its seems a very "dumb human" can generate a near-infinity (not truly infinite, for the average lifespan in the modern world is some 75 years, or so, and with 16 waking hours * 365.25 days* 60 mins/hr*60 secs/min * X# of thoughts/second; you get the point!) of sentences, given a very small vocabulary, some spare time, and an extremely simple syntax. Furthermore, despite the simplicity of the syntax, REALITY is complex, so, almost invariably the human animal--if he/she lacks a syntax complex enough to describe said reality--finds a way to discuss those matters which must be discussed.

So I don't think "game [ing] the metric..presented" is the point. The metric is a theoretic limitation and a "guideline" and an insight into what is occurring in the "mind"--loosely defined--that is considered "intelligent".

I am not certain that Strong AI will be EXACTLY like Human Intelligence. I am certain it will be close, but since I can't predict what any one person will do under a given set of circumstances, I don't expect that AI will follow my expectations for intelligent actors either! (I read a psychology study about this once; it turns out people are surprisingly bad at "intuiting" the behaviors, emotions, thinking and reasoning of others).

You would be right to say that I am proposing to "brute-force" the "human intelligence algorithm", and, thus, to simulate human intelligence. I think it is already happening, but not yet to an extent where we can declare Strong AI to be a reality (e.g. Big Blue, Watson, Siri, etc.).

So, what about light computers? I woke in the middle of the night (4am) and dreamed up this device; Blue Ray drive with multiple read heads! Feasible? I don't see why not! How about ten such drives, an eleventh always loading a new disk while the other read, and all of them hooked to CD changers (100 to 200 disc CD changers available, used on ebay!). Optical jukebox/RAID array!

My only REAL questions are (1) Can the photodiode step be skipped, and the blue ray signals be piped (i.e. lensed) right into some fiber optic strands, and (2) what does one do with 80 Gigabits (of optical data) per second?

Here is a fanciful (but do-able?) schematic!

Whoops; jpegs appear to be better than TIFF (and Bmps not good at all?!).

well found you a picture 6502 ^%)@

Your bender needs a retina! Here's a ping-pong ball, and some cardboard. Leonardo likes it!

https://en.wikipedia.org/wiki/Leonardo_ ... yhedra.png

https://en.wikipedia.org/wiki/Petzval_field_curvature

In just another 16 years, we might get a president that bothers to fund the sciences, and then I can build one that isn't made of cardboard and ping pong balls!

Not bad SMT soldering; a project of mine, but the muscles are weak and the hand is flimsy!

Next time I'll try (1) thicker Nitinol musclewires, or (2) stranded muscles wires (though I expect shorts or trouble with heating and cooling cycle /times or (3) graphite tube fingers, with teflon joints (instead of paper), (4) antagonistic muscle setup, and maybe "tendons", and/or (5) some kind of flexible, non-conductive insulator (oxymoron) around the wires, so that they might be bundled and have their forces additive, while not retaining heat or shorting from one wire to another.

https://www.youtube.com/watch?v=M7jQEji ... e=youtu.be

And so, my robot eye fantasies continue. A paper caught my eye today, whilst shuffling my mess from oneend of the room to the other. Le Scribe Accropi! The first schematic eye ever. Ancient Egyptian statue. Copper "retina", so to speak. Painted iris. Lens, so that, possibly, when tomb raiders torchlight flashed off the back of the retina, they would leave in a fright, and leave the pharaoh to his treasure (sp[eculation, of course!).

Jay Enoch is the researcher. He has been involved in vision research since the 50s and 60s I think. Last I checked, still a prof in California somewhere. I am surprised that few people seem to know the history of lens design, as laid out by Rudolf Kingslake in his book of the same name. The Petzval (Coddington) curvature was significant obstacle to the proper development of the photographic lens, and took about a decade or two to properly fix. It seems to have been forgotten, that we have developed methods to flatten the field. Now that we have a suitable photosensitive surface, even capable of a rough sort of 3-D shape--if one tries hard enough--nobody seems willing to re-un-design it!

And what of the fundamental knowledge of Ptolemy, al Hazen, and subsequently, Snell? Why would we (modern day researchers) ignore what is known about optical materials, namely, low refractive index materials? The human cornea, lens, and aqeuous and vitreous humors, have relatively low refractive idices, compared to most optical glasses. The plastics are a bit closer, but, truly, one must gravitate towards optical crystals in order to mimic the natural materials. Unfortunately, those crystal materials are either harder to work, or exhibit some sort of objectionable, nonlinear properties (birefringence, polarization, etc.).

One of my lens design books has a paper model, with some ray tracing through the human eye. But I can only find one reference to anybody who has EVER tried (besides the sculptural attempt of the Ancient Egyptians) to construct a working, 1:1 scale model of the optics of the human eye (well, actually, his was more like 1.5;1; a smidgen bigger. G.H. Gliddon circa 1927; Dartmouth). As for "silicon retinas", under Carver Mead and the neuromorphic research programs, several researchers have attempted some implementations. T. Delbruck, for one, Misha Mahowald, another (I just stumbled across the title of her book, and am kicking myself for not obtaining a copy from somewhere! "An Analog VLSI System for Stereoscopic Vision".). Boahen is another who goes the Neuromorphic route. While I do not favor "chips that act like neurons" as an optimal research strategy, I should mention that I see its appeal, as does that DARPA guy (Robert Colwell), whose you tube talk https://www.youtube.com/watch?v=JpgV6rCn5-g I've referenced on another thread.

Other "retina-like chips" include the first, van der Speigel et. al., circa 1989, wodnicki et. al, and sandini et. al. There are many more, and I have only scratched the surface. I know enough to know, that if the device I am thinking of exists, it sure is hard to find.

So, on the "biology" side of things, I could mention Ramon y Cajal, winner of the Nobel Prize, and his fine microscopy and artwork of the retina (and his evidence supporting neuron theory!). Osterberg, the first to estimate the number of retinal cells, bears mentioning, though the numbers (circa 1990s) are a bit reduced by modern statistical sampling (Curcio et. al). Also, Gullstarnd, another Nobel Prize winner, should be mentioned, as well as Polyak, the U. of CHicgao prof. who wrote several large, seminal tomes on the vertebrate and human eye.

The modern textbooks that bear mentioning are probably, Clyde Oysters "Human Eye; its structure and function" and Smith and Atchison's work on the dipotrics of the human eye. Any person who winds his way through these works and then proceeds to tell me "why make a robot eye. its impossible. its already been done. we already know everything about the human eye.", is just showing their ignorance.

Other fine works that bear mentioning are Nicholas Wade's "Natural history of vision" and David Parks "Fire in the eye". And just two summers ago, I got a chance to go to the national archives and view the Harvard optical Laboratory papers on "Optical Fluorite for Aerial Lenses". Quite a journey for me, and not something I take lightly. One hopes my fluorite doesn't crack, when I start working it. Thermal stress is easily generated and the crystal doesn't hold up to well. Also, its tough to polish (so I've been told).

Still have much design work to do, on my student "OSLO" CAD program. I'll get this lens thingy made!

Well, that's it for tonight.

Whoops, one last thing. If this circuit doesn't work, at least its pretty (in the original colors, it was nicer, but this is just black and white).

It was 9, KAC-9628, CMOS image sensors, 48 PLCC/CLCC. All SMT parts. Kind of hassle, so I never finished it. I give myself 4:1 odds that I don't ever get that particular project working. But I might have learned enough to make my next 9 image sensor project function.