Slashdot Mirror
$6 System-On-A-Chip Mimics Human Vision
Brian McLaughlin writes "This article in TechWeb describes a Visual Perception Processor (costing $6) that can automatically detect objects and track their movement in real time, according to Buereau d'Etudes Vision (BEV). They claim that a full-blown vision processing system/application could be built for less than $50 that rivals current state-of-the-art $10,000 systems. Sounds pretty cool.
" Heck, with my vision, I could tear my eyeballs out and simply use these, at a fraction of the cost of new glasses.
I wonder if this is covered under my HMO. :-) Seriously, though, I wonder about the insurance industry and how they are going to handle things like this coming up. LASIK is getting more and more popular, but it is still expensive. If getting a processor like this is cheaper than LASIK and cheaper than conventional glasses -- what's the future of vision plans? At what point do they spring for the implants/glasses/lasers to fix you up? Of course, with all the genome stuff going on, maybe they'll make you fix it before the child is even born.
Uh oh. If AIBO implements this technology, I might have to buy one after all!
ICQ: 49636524
snowphoton@mindspring.com
Got Rhinos?
Heck, with my vision, I could tear my eyeballs out and simply use these, at a fraction of the cost of new glasses.
Hrm... I don't know, but for some reason that just dosn't sound like a good idea to me...
[ c h a d o k e r e ]
ReadThe ReflectionEngine, a cyberpunk style n
I wonder if we could add in a zoom lens, or maybe x-ray vision (Like in the latest Bond movie).
$50 for a reactive vision processing system? Couple that with a cheap (or free) reliable operating system, and cheap networking that has lots of addresses (IPv6), and you could put watching devices on every street corner, heck in every house (you know, for the safety of the children!).
Goodbye privacy.
George
A) does it run Linux and
B) if so, will a beowulf cluster of these things make bugs infinitely shallow?
Great, now if I poke out his eyes, he can replace them and still find me. Plus, he will be pissed off at me for poking out his eyes, so he will eat me very slowly.
...if you'll excuse the awful pun.
Seriously, though, this has some seriously cool implications if it actually works. It reminds me of a story posted a while back on Slashdot, about a blind man who'd had prototype cranial vision implants back in the 70's. He's currently got some semblance of vision (truly achromatic, poor resolution, and about the size of a 3x5 card held at arm's length, but vision nonetheless). It's had some upgrades over the years, but it's still ancient technology by modern standards. What would happen if they could somehow retrofit that implant with one of these, assuming it worked? While I doubt it'd be anything near human vision, it'd certainly be better (both in terms of vision quality and, probably, physical bulk) as the system he has now.
It'd certainly be an interesting one to try.
Driving down highway
Blue screen of death suddenly
I crash into tree
If this chip is really as capable as it's made out to be, it will mean a great deal to people who are primarily interested in autonomous mobile robots, as opposed to computer vision.
I could imagine hooking something like this up to a pioneer and solving a bunch of problems.
Sort of makes me wish I were still a student, with the time and resources to play with robots...
I could well believe it. Several months ago I had lost my glasses. Now, I didn't think that my vision was THAT bad, and I thought that maybe I could function without them. However, not fifteen minutes after leaving my house I walked straight into a truck. I mean, it's a common mistake, those trucks are pretty tricky. You can't notice them unless you look REEEEAAALLLY carefully. And it wasn't my first incident with a truck either. I can't count the number of times I've been stalked by Log Trucks.
Now, you might think that's a joke, but Log Trucks are a SERIOUS problem. Once one has its mind to get you, get you it will. They are VERY ruthless, and will stop at nothing to leap out of the woods when you're not looking and have thier way with you. But don't take my word for it, read the archives at the alt.fear.logtrucks newsgroup.
At any rate, I fell onto a rake after I walked into the truck. Poked out both of my eyes. But on the slightly less bleak side, at least I didn't cry (I couldn't; I had no eyes, or tear ducts. Rakes suck). Anyway, once I got a hold of myself, I asked a passing mime to direct me to the nearest hospital, and I hooked up with a vision unit that cost me $15,874.97 plus tax. To be fair, that DID include the surgery, so I don't think it was that bad of a deal.
Anyway, the vision isn't really that great. I mean, what I see is clear, and I have a nifty zoom option. However, I have lost the ability to see clothing. EVERYONE is naked. Oh, yeah, make your perverted little jokes. And yeah, it was neat at first, but let me tell you, you watch ONE episode of the Jerry Springer show and you'll change your tune REAL quick. Let alone going to grocery stores. I don't even want to THINK about going to a thrift store now.
Which brings up another issue: wearing clothes. I can feel them, so I know when I am. And just because I look naked to me, doesn't mean I want to look naked to everyone else. However, because I can't see what I'm putting on, I end up with some pretty strange outfits. I can't really tell if I'm wearing a three piece suit or a "Hello Kitty" sundress. The only indication I get is the trail of giggles. Do I look THAT weird in a suit?!?!! I guess so.
Anyway, I hope whoever is developing this system has read my story, and makes sure this bug does not persist. Oh, and will someone tell me where the next Linux convention is so I can steer right the fuck clear of it
If I understood the article corectly, and I rarely do. Then someone can use this $50 or so chip to build devices such as: motion detectors, better robot controls, new space craft for NASA to crash, and the one I really like a car that drives itself without getting blood on my bumper.
Devil Ducky
MY peers would get out of jury duty.
Sony has sold video cameras for years that could track people. They are used all the time with room video conferencing so you can have it track whoever is presenting in a meeting. They cost about $1200, I've got one sitting in a box in my closet right now...
"Gait identification"?!?
This is possible? I guess it would work on the Fat Albert & the gang, but the Minister of Silly Walks would fool this thing faster than your machine can go "bing"!
This guy already has a functional Visual Cortex, what would be the point of replacing it? What this guy needs is better/ more connectors in his brain, not a new one.
[ c h a d o k e r e ]
ReadThe ReflectionEngine, a cyberpunk style n
my first post, be a crap. signal to noise ratio = 1
Damn you! Poke out MY eyes will you? Eating you slowly is too good. I will eat your sister in front of you. I will let the mime eat your mother. You will watch, and because I am the clown, I will be funny. You will laugh and scream and scream and laugh. And then the mime and I will eat you. This we will do for you.
-the clown
Cool, I really want to see you in a Hello Kitty sundress. With pigtails, white lace stockings and black pumps. Now there's technological progress!
The brain-eye system uses layers of parallel-processing neurons that pass the signal through a series of preprocessing steps, resulting in real-time tracking of multiple moving objects within a visual scene.
Get it now? So please guys, don't get any *SMART* ideas and jab your eyes out so you can buy a nice new $6 dollar replacement. (I know no one was actually serious about that, but you guys did still miss the point. Its not the camera thats important, and it would do no good to put xray or a zoom lense on there because you're not looking for the output of what its seeing. Its a TRACKING device. So its not used for monitoring either... geez.) \rant
I am a leaf on the wind. Watch how I soar.
The GVPP's major performance strength over current $10,000 vision systems is its automatic adaptation to varying lighting conditions. Today's vision systems dictate uniform, shadowless illumination, and even next-generation prototype systems, designed to work under "normal" lighting conditions, can be used only from dawn to dusk. The GVPP, on the other hand, adapts to real-time changes in lighting without recalibration, day or night.
Okay, so what they're claiming is that their brand-new, $6 ($50 in total) device can do things which long-standing scientific projects costing $10,000 cannot? Am I the only one who thinks that this sounds somewhat fishy?
The GVPP was invented in 1992, when BEV founder Patric Pirim saw it would be relatively simple for a CMOS chip to implement in hardware the separate contributions of temporal and spatial processing in the brain.
Again don't you think that all of the many computer scientists and neuropsychologists working on machine vision wouldn't have thought of this themselves? I've read a fair bit on the theory of vision processing and pattern recognition and it's a hugely complex subject. And now a small research company has cracked it? I don't think so. If you read the list of things which they say it can be used for it comes across as being a huge gimmick - they seem to have listed everything they could think of that might be worth money.
All you need is one extra chip to hold the banner ads.
When you are dancing with wolves, never limp
automatically detect objects and track their movement in real time, according to Buereau d'Etudes Vision (BEV). They claim that a full-blown vision processing system/application could be built for less than $50 that rivals current state-of-the-art $10,000 systems. Sounds pretty cool It looks like we're in for a slew of good-and-bad things to come. This silicon eye is so cheap that we'll end up seeing it installed in the oddest of places. Everything from the (1984-inspired) eyeball-in-the-TV (gives new meaning to the CBS logo) to anti-collision systems in vehicles. For $50 you can put one on your kid and see what they do (Here, Honey, it's a pendant from Aunt Huxley). Worried about which neighbor's pooch is pooping on your lawn? A few modules can be used to watch and track it home. I guess I'm a pessimist. I see more bad things, particularly in the area of privacy loss, than I see good.
"First things first, but not necessarily in that order."
- Doctor Who
Damn cut and paste!
has anyone noticed that this might make implimenting a visual system for robots easier. heck, maybe in a few years i will have that robot maid who can take care of all my needs.:-)
Beer Die is the game of champions Learning To walk my own path.
Alongside a CMOS imager on its 2-by-4-inch evaluation board, the GVPP has been demonstrated as capable of learning-in-place to solve a variety of pattern recognition problems. It has automatic normalization for varying object size, orientation, and lighting conditions, and can function in daylight or darkness, BEV said. A complete GVPP system, including the charge-coupled device and all support circuity, should cost less than $50,
I don't get it? Does it use a CMOS imager or a CCD? The two are NOT the same. CMOS imagers are inherently digital devices, based on the same CMOS transistors powering any digital IC (including microprocessors, etc..). On the other hand, CCD (charge-coupled devices) are inherently analog.
Most (if not all) digital cameras and other imaging devices are CCD based. However, CMOS imagers are starting to evolve, and should be cheaper to produce.
Check out http://www.photobit.com
Thanks for making this point. It's still damned cool that they call pull out all this visual information, especially in such diverse lighting conditions. A university pal of mine tried to do some vision stuff a while back - I helped a little - it was very difficult to identify objects in a scene. Even when it's, eg, a red coke can in a white/light scene. To have such a cheap chip be capable of identifying and tracking 8 objects in a scene (presumeably the sensor is outputting a pixelmap) is incredible.
This system does not see the same way that humans do, nor is there any mention in the article that you could actually get a picture out of one. It's designed to locate movement within its field of vision...nothing more. It doesn't notice, and probably can't 'see', things that do not move.
I wouldn't want to replace my eyes with these things. I'd be able to see the animated banner ads, but not read the rest of a static web page!
Genocide Man -- Life is funny. Death is funnier. Mass murder can be hilarious.
with an array of these, i bet i could keep up with all the porn sites on the web.......
After the prior Dune article, I had to mention this (these). Someone help with the spelling!
Anyone know where I can get a hold of one of their evaluation boards? Back in 1997 they were offering the boards for just under $1k.
Actually it's not as impressive as it looks.
They say that The $6 Generic Visual Perception Processor (GVPP) can automatically detect objects and track their movement in real-time, according to Bureau d'Etudes Vision (BEV)
This can be easily accomplished by a technique called blob tracking - which is the coarsest image vision technique. A similar project (impressive too) was developed at a japanese company (although I don't remember exactly where). It was some sort of interactive game where a pet was playing with you in a projected image. You moved, played with it and it seemed to understant if you touched him,petted him etc. The catch was that the camera was filled with cameras and they detected the movement of your hands. Given the speed, direction etc. they could actually appreciate what you were supposed to do.
Cute, but nothing interesting from a research point of view.
I assume that they are doing the same. It is very easy to identify the movements in an image (you can do it in real time even on a Pentium). Check any image vision book for details.
Probably they built some chip that works at the speed of a controller (i.e. very fast) but, as any controller performs very few operations.
Still they don't say anything about actual image understanding.
And this is where comes the commercial part. Because they actually are not saying their chip can understand an image. They simply can track motion. That system wouldn't have a clue whether that movement is a fighter or a flying orange.
It may be useful in a computer that is used in a vision lab, but we're quite far from industrial pattern recognition or image understanding.
So please don't take commercial ads as truth
If you are French please don't read what will follow.
After all, they are French, and they are the best sellers in the world. Every one believes that French wines are great and French women are beautiful. Have you ever tasted those sour poisons ? Or ever went to France to watch their women ?
Just did a quick search, and found out that GVPP isn't exactly new (the article mentions that it was invented in 1992):
http://www.techweb.com/wire/news/1997/09/0913visio n.html
Seems the price has gone down "a bit" since '97 though:
The modules measure 40 mm2, have 100 pins, and can handle 20-MHz video signals. The chip is priced at $960. On a card with a socketed GVPP and 64 kilobytes of Flash RAM, the price comes to $1,500.
$6 sounds much better to me :)
This sort of thing is good news for people with seriously degrad(ed/ing) sight like me. Right now I spend nearly $400 on lenses right now before I even start looking at frames.
The technology isn't good enough for full vision replacement yet but it's only going to improve. What I'd like is 'augmentation' applications:
- 360deg vision
- microscope extension that can be
put into a machine and get a good
view of the motherboard or components
- direct output from a video game or computer
*heh*
Sign me up.
--Ruhk
404 Error:
This was the first thing seen with the new device.
Thank you.
You're more or less correct in your analysis of the chips, but come on, lighten up! Part of the fun of reading /. is coming across off-the-wall musings people post.
Try to think out of the box a bit more, some of the most innovative inventions come can come from seemingly stupid ideas - why do you think that nothing is out of bounds in a brainstorming session - ideas spark ideas.
Life sucks but death doesn't put out at all....
They aren't talking about doing object detection or segmentation, much less tracking, in a rich optical environment. Doing segmentation and tracking in a sparse and controlled visual environmant might useful in a factory environment, but it is going to be of very little value outside of that realm. That means that this chip is much less than it appears, and, frankly, isn't even all that new. Go look at Carver Mead's work in the early and middle eighties, if you don't believe me; it could do the same thing for the same price. Heck, go look at Eric Schwartz's work in the late eighties; it could do much more...for much less.
Seems no one noticed your clever haiku yet.
I knew I shouldn't have squandered all my moderator points yesterday...
I misread the techweb article. If anyone is intersted there is another (slightly more technical) article here
2^5
Hrmm, for robotics enthusiasts, has anybody tried a fisheye vision system? Could be some interesting stuff :-)
Eh...
You can find more information about the chip at http://www.eetimes.com/news/97/9 71news/vision.html. This tells a little more of how the chip actually works. provolt "I joined the giant collective brain and all I got was this lousy post.
This sure sounds like a technogical device that could be used to circumvent access control.
If you live in USA, don't hold your breath waiting for these.
---
As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
That's right kids, when you wear the new PokeTiara you can find all of the PokeTrainers in the area. Then you can have big Pokefights!
Never mind the fact that one of these imaging things is in the Pokecrystal on the PokeTiara telling Nintendo whether Pokemon is played more in little groups or big groups. It can even be modified to tell teachers if Pokemon is being played at nap time.
The idea may be crazy, but I'm sure real company executives (the gov't wouldn't try it when it's cheaper to buy the information from the megacorps; especially after the lawsuits) could come up with crazier ideas.
Is this post not nifty? Sluggy Freelance. Worshi
Kuro5hin also beat slashdot to news about the XFree 4 news.
...if the general-purpose hardware wasn't so stupid. Of the millions of transistors on a modern chip, most of them are wasted in maintaining the illusion of sequential operation, while the OS writers go to considerable trouble to create the illusion of parallel operation.
Furthermore, there are the huge (in terms of transistor count) banks of flip-flops which just sit around most of the time, and the costly layers of cache all working their hardest to maintain the illusion that it is RAM. Meanwhile, software optimizers make sure to access memory sequentially to avoid upsetting this illusion, which would ruin the performance.
You can justify all this nonsense with the argument that software is written for sequential machines with RAM. It's a circular problem. If somebody would just release a cheap massively parallel system, the programmers would learn to use it efficiently.
You can make a complete processor in a few thousand transistors (as this guy has done, though he goes a bit off the deep end...), and you can add a bit (a few K) of high-speed RAM and network them easily enough to make a (dare I say it?) Beowulf cluster on a chip. Each might only run at one tenth the speed of a modern CPU, but you could have hundreds of them for the same cost, giving you bips and gflops for the price of mips.
It would also make the whole design process a lot easier and faster. One simple processor, repeated hundreds or thousands of times. Every advance in production would bring a direct and proportional improvement in performace, with a tiny added design cost. Forget special graphics or sound processors, just plug in more processor banks like you would add memory today and watch your system fly.
C'mon hardware guys, we software guys aren't that stupid! We don't need your illusion of a 386!
>Heck, with my vision, I could tear my eyeballs out and simply use these, at a fraction of the cost of new glasses.
This is to you, Hemos, if you're reading this. I've had it up to HERE [indicating invisible area above forehead] with your half-witted little comments at the end of your posts. If you don't have anything even remotely intelligent to say, just don't say anything at all, and spare us both the trouble. Your clever little quips are neither clever, nor humourous. Try to save a little face every once in a while.
If you don't understand why your comment deserves such derision, just think about this: It's a vision PROCESSING system. Your eyeballs are not vision PROCESSORS, they're the biological equivalent of CAMERAS. The front part of your brain is where the processing takes place. Therefore, putting a PROCESSOR in your eyesockets will not help you see better at all. I do imagine that replacing a part of your brain with it would help, however.
Also, this isn't very appropriate. We're not your buddies... We don't know all about your horrible eyesight. I'm sure you've got terrible eyesight, but there's no need to share it with us all. Try saying something relevant to the article. For example, "Those French sure can make some nice chips!"
I really thought that, from a large majority of the slashdot community, this needed to be said. Please try to think before you speak. If you can't do both, do neither.
Thank you.
---
Epi "Flamebait" Taph
I'm sure they've done some great work but my personal belief is that it's been subject to a marketing department's hype machine. The vision system in a primate is one of the most complex parts of the brain and is still not entirely understood by neuropsychologists.
There are three main pathways from the eye - the magnocellular (connected to the rods mainly and used for brightness and motion detection), the parvocellular (connected to the cones mainly and used for colour determination) and the koniocellular (whose function is less well known). These three pathways feed into the V1 area which acts as a feature detector, then into the V2 area, which detects colour features and movement and then into a variety of different areas including the V3 (shapes), V4 (colour) and V5 (motion and positioning) areas, the parietal lobe, the thalamus and the various inferotemporal and interparietal areas among others still being found.
All of these different areas seem to have some bearing on vision in its entirety, and show just how complex vision is. As such I think that any claim that a company has suddenly perfected a chip which allows complex visual capabilities is suspect until hard facts and experimentation can prove or disprove the claims.
Let's not throw our eyes out yet. I would kind of like to see more than eight things at a time and I don't think I will have the luxoury of telling the chip what sort of things I might want it to monitor. I know you can string multiple chips together for a greater number of items monitored but the key here is that they monitor pre-defined objects. They couldn't possibly serve as an eye replacement. This is particularly important to me as I have a disease that is detroying my retina and choroid (Choroideremia, which is a special form of Retinitis Pygmentosa).
While this won't be letting us get new eyes, one neat thing would be to add this onto a wearable so that one could automatically track things in your field of view(anybody for a game of cups?). Besides defeating sleight of hand, it could let you do the "glowing trail" effect that is used in hockey games and the like(automatic bullet tracking for the SWAT team as well).
And of course, my favorite, Predator style targeting reticules;-,
There is a much better article here from September, 11 1997. Does the 1999 article say anything new about this chip?
(There's also a blurb on the media demonstration here from the Nov. 17, 1997 Japan Times.)
-ac
On the other hand, this could also be the basis for technology that tracks where you go and what you do. Under the auspices of controlling crime, criminals could be "flagged" and watched, traffic policing could be automated, etc. Where it gets scary is in who determines what suspicios behavior is, or who qualifies as needing to be watched, or the fact that you are removing the human element from the decision making process of evaluating a crime.
In the end, both the citizens and the govornment want this kind of pervasive, intelligent, monitoring technology to be ubiquitous. The difference is that citizens want to be able to turn it off.
You can make a complete processor in a few thousand transistors (as this guy (Moore) has done, though he goes a bit off the deep end...),
(Off topic, but...) Oh, the FORTH chip guy. He's wierd, but very competent. He's into things like generating video signals in real time, in software, to avoid needing a video chip. He used to use an interface with only three pushbuttons as input (no keyboard), with which, by suitable manipulation, you could not only operate a menu system, but program in FORTH. It's worth a look just to see how far minimalism can go. It's not all that useful, but if you architect systems, it's worth seeing his approach.
If MS were to power this type of thing, imagine the benefits: - Blue Screen of Death in lieu of vision at the most inappropriate times - Buggy, inefficient code that would not allow your vision to refresh. Instead, you get a box of unrefreshed (invalidated) vision in certain places until the repaint call is serviced - Other services popping up for attention instead of staying in place (ex. Yes/No dialogs) - Wonder what "Critical Updates" would look like? - Wonder what bug reports from MS would look like? (This issue only affects a small minority of users. It is not destructive...) - Oh, and if you don't register in 120 days, it shuts down until you buy a new one, reinstall, or register. That's all I have to say about that. M
Check out the failure of the Connection Machine and you'll see that actually us software guys are too dumb to write good massively parallel software. I do it, and it's hard, especally the debugging, and that's just parallelizing to 64 processors. Perhaps there are other paradigms other than the standard imperative programming languages that would make using these architectures more efficient, but as it stands now, it is very hard to effectively use parallel architectures. With the introduction of Intel's Merced, a VLIW architecture, we will see what happens when excellent compilier writers take a crack at a somewhat parallel architecture. The problem is not as easy as you make it out to be or it would have been done.
All articles appearing in Kuro5hin.org is appearing in Slashdot some hours later
He st uck a whole computer in a mouse</a>, apparently just for the hell of it. Knowing him, it probably wouldn't cost significantly more than an ordinary mouse, either.
Which isn't exactly new... tho c64s and amigas used to have quite nice video hardware for their time, it was not unusual, esp. on the c64 to get very close to creating the video signal in real time (which allowed one to crate things which went far beyond the original specs in things like multi color support (adding completely new colors to the hardware ;-), higher resolutions and drawing in screen areas where according to the documentation the hardware could not create screen content.
Can I get mine outfitted by the same designer who built that Sony camcorder that looked through clothes?
it was done in the 1970s
[ c h a d o k e r e ]
ReadThe ReflectionEngine, a cyberpunk style n
It seems very dubious. Perhaps their parallel system will allow tracking to be done more cheaply and quickly than existing systems, but tracking and vision in general are still, at a fundamental level, very much unsolved problems.
The human visual system uses countless contextual, knowledge based cues to make sense of the world (e.g. take a look at some optical illusions); even a simple task like tracking requires a lot of background knowledge. To claim that their mimics the human visual system is, frankly, unbelievable.
I believe this is one of the first steps to a world similar to what William Gibson describes in his novels, where humans and technology interface in a whole new way. With these types of advances a whole slew of ethical and privacy issues will develop, as they do with any revolutionary technology, but I believe that in the end the benefits will far outweigh the drawbacks.
Now all that is needed is for some other emerging thechnologies from varied fields such as nerve re-growth, computer miniturization, etc. will combine to produce things like relacement/upgrade body parts.
Personally, I can't wait for the first brain-embedded computers to emerge. I'd be willing to be one of the first human testers of such technology.
Any comments/ideas?
If it is really capable of what the designers are claiming, it will be really cool.
Lets just hope the military or the government don't go and buy it up; if they did, it would disappear from the public domain completely, and forever.
I am not sure this system would contribute to loss of privacy. From what I can make of the article, it is just a motion tracker. It wont pick your face out of a crowd and alert someone.
It is already possible to monitor people with hidden cameras. This system could only track someone's movements.
What happens when the target turns around, obscuring his/her face, or whatever part was being tracked? Will it compensate by locking onto a different part of the object/person? What happens if the target becomes obscured?
I guess it could be used to follow someone walking down an empty street.
The article makes this system sound very capable, but it would be desirable for it to be able to compensate for the above problems (I doubt it would be able to deal with all of those mentioned) for it to be useful to follow someone automatically, anywhere they go.
Gotta wonder, a couple of these operating in tandem would give you a great depth perception, visual coverage of movement... wonder if it would make for more accurate smart bombs and ABM devices... Ronny Raygun Rules!
Our fight is not against flesh and blood, but against the rulers, against the authorities, the spritual forces of evil
Congradulations! You just explained almost exactly what IBM is doing with Blue Gene :). Blue Gene, if I recall correctly, is(will be) made of 2cmx2cm chips with simple little RISC CPUs (somehow 20 per chip, probably not nearly as simple as the aforementioned CPU), having many chips on a board, and many boards in a rack, and link it all together somehow and you have a system that can handle a couple petaflofs/second. I think the biggest problem in making it work, however, is how to get the millions of little processors to work well in paralell. The system is specced to have something like 4TB/second of total bandwidth, but it just seems too much like a massive Beowulf cluster to me to be efficent. If hardcoding for 2 or 4 CPUs is difficult, coding for 2 million CPUs will be your worst nightmare, unless IBM somehow makes some amazing software that Blue Gene runs on itself to direct every little instruction.
This Yahoo story says the auction of the technology is under way now.