Hitachi Develops New Visual Search

Tech.Luver writes to tell us that Hitachi has developed a new visual search engine that can supposedly find similar images from within millions of video and picture data entries in around 1 second. "The technology assesses the similarity of images based on image characteristics presented as high-dimensional numeric information. The information is acquired by automatically detecting information regarding the images, such as color distribution and shapes."

Comments

Hmmm.... robotics?

[fyngyrz]

This is interesting to me - if it performs well - because this is one of the key missing elements for robotics; robots have a lot of trouble trying to match the environment around them to stored records of objects unless the environment is severely constrained. I'm not speaking of AI here (or at least, not yet) but just robots that would be able to clean your floor, carry your groceries, navigate in a burning building, walk your dog, tend your lawn. If they can classify images against stored images well, we're that much closer to generally useful and at least semi-autonomous robot devices.

Training might be a little annoying the first few times, but once you had a good database, you could replicate - or share via RF, that'd be freaky... neighbor's robot learns what a ferret looks like, now yours knows too - so that newer models were more and more informed right out of the box. Crate. Coffin. Whatever.

Add an associative database so that images normally found near other images which have just been found are searched first, and perhaps you could get the general search time down from the quoted 1 second, I'm thinking. One second is kind of pokey for a lot of robotic applications. But if the thing is in a kitchen, why would it need to be looking to recognize images that are found in a shipyard?

And I, for one, would welcome our semi-autonomous, environment recognizing, floor cleaning robot underlings.

Re:Hmmm.... robotics?

[dotpavan]

what I am curious about is the computing resources required to process.. other than the algorithm, is this one of the reasons which is delaying the emergence of search in the field of images/music/video on a commercial level? riya made some strides, but is still "learning"

Re:Hmmm.... robotics?

[lawpoop]

I'm not speaking of AI here (or at least, not yet) but just robots that would be able to clean your floor, carry your groceries... Well, you are talking about AI here. It turns out that it's relatively easy to make a computer that can beat humans in chess or do complex math equations, but something as simple as walking with 6, 4, or two legs, which a lot of really stupid organisms do, is really difficult. Something like distinguishing 'indoors' from 'outdoors' or a cloud bank from the bushes, seems way in the future.

My pet theory is that we don't have the right kind of device yet. A mind, the 'function' of an organic nervous system, is not a Turing machine. I don't really understand the math behind it, but Goedel's incompleteness theorem seems to show that a human mathematician can understand certain mathematical proofs that a Turing machine can never prove. Since all computers are a essentially a Turing machine, no matter how fast or parallelized they are, or how much memory they have, they will never be able to do what a human mind can do. So, maybe someday we will have artificial intelligence, or, a floor-washing robot, but we currently don't have the right kind of device that can do it.

Re:Hmmm.... robotics?

[moderatorrater]

There's got to be some algorithm that our minds use to determine whether something is indoors or outdoors though. Why not gather data on how humans do it and then replicate that with AI? When that's done, move on to another concept until the pattern of how humans recognize things emerges, then copy that for new things.

Re:Hmmm.... robotics?

[Anonymous Coward]

The problem with this idea of storing images is that it could be pointless to store that much data. Your brain does not store images of say a tank. Your brain says "Does it have tracks?", "Does it have a turret?", "Does the turret rotate?". Well to figure out if it has tracks it would then search characteristics about tracks, then search for information about turrets etc. Once that is done you know it is a tank, now you need to know what kind. So you would need to know stuff about specific types of tanks like the Russian T-34 has a slanted front. Maybe search for a Russian flag, well then you have to figure out if the symbols on the tank are a Russian flag etc.

Most everything is made up for basic shapes. Some of those shapes may only be described by mathematics like for example, circles, triangles, squares etc. Now you just have to save how those shapes fit together to make a bigger shape but a shape in 3 dimensions.

Say for example you wanted to find a trash can. You would look for something that looks like a cylinder or a rectangle with a hole in the top. It would then have to be sitting on a flat surface. The top of the can may not be well defined due to the bag. Thus you have found a trash can. So rather then searching for images search for common shapes and combine those shapes to find what you are looking for. The problem is picking out the shapes.

Re:Hmmm.... robotics?

[fyngyrz]

Well, you are talking about AI here.

No. I'm not. There are robotic vacuums and lawnmowers right now. I'm just talking about giving them some eyes so they know not to mow your puppy or your child or your roses, or vacuum up your engagement ring. Teaching a robot firething not to step into a hole in the floor, and to rescue people before pets, and pets but not stuffed animals.

It turns out that it's relatively easy to make a computer that can beat humans in chess or do complex math equations, but something as simple as walking with 6, 4, or two legs, which a lot of really stupid organisms do, is really difficult.

Not so difficult that it hasn't been solved multiple times, multiple ways, including such variations as stair-climbing and running. Nothing to do with AI, either; just a progression from over-complicated attempts to solve using complex equations over the whole assembly to simpler approaches like fuzzy-logic based feedback systems that work right at the joints.

Something like distinguishing 'indoors' from 'outdoors' or a cloud bank from the bushes, seems way in the future.

Not if there's a good image-matching mechanism, it isn't. The concept of "looks like" is a very powerful one. That's what Hitachi says they've done here; we'll see if it lives up to the report.

My pet theory is that we don't have the right kind of device yet.

Keep that theory warm. Reality has a way of bringing the cold, fast and harsh. My pet theory is that a serial computer architecture can emulate anything, anywhere, given the proper code, enough storage and enough time to jump through all the hoops; to which I add, once you get it working, you can optimize the code and the hardware to do the job better until it is in the realm of the practical, if the investment is worth it. And for AI, IMHO, any investment is worth it. That's been the history of every solved problem so far, and I see no evidence that any solvable problem will be any different. And intelligence is solvable; after all, nature solved it may ways.

Re:Hmmm.... robotics?

[dj_tla]

All of the robotics problems you described - cleaning your floor, carrying your groceries, navigation, etc. - are AI problems. In fact, even the "(1) high-speed visual similarity search using two-step search clustering technology" that Hitachi is promoting could be considered AI. You see, in computer science, if a problem has some kind of real world application (even in theory) it goes under the moniker of AI, and this gets papers published. In my opinion, AI's just a rebranding of the rest of computer science; object orientation becomes "frames," complicated parsing problems become "natural language processing," robotic navigation is "pathfinding." But I digress!

Anyway, more to the point, Hitachi's technology doesn't do any learning, and thus, doesn't require training. It's also not particularly new technology, they abstract out some features for the images in their database, then search for images with similar features. This is probably oversimplification, I'm not a computer vision guy, but I'm relatively confident that Hitachi has not furthered the state of the art in computer vision. The reason I guess it's newsworthy is the second part: "(2) faster reading through optimized data allocation on an HDD." This isn't a new technology either, though. I'm not sure if TFA is really that newsworthy.

Robots have a lot of trouble trying to match the environment around them to stored records of objects unless the environment is severely constrained. This is mainly because robots are specialized. For navigation, infrared/laser/sound sensors are better suited, as they tell the robot how far away obstacles are. Robots that are concerned with identifying objects usually do not need to navigate.

Re:Hmmm.... robotics?

[kebes]

Your implication is that the human mind cannot be reduced to a Turing machine. I am in the other camp--who believe that the mind is subject to rigorous physical law, and that physical law can be expressed arithmetically (in principle), and so the human mind is a Turing machine.

Godel's theorem says that a consistent arithmetic system will contain unprovable truths. Put otherwise, such a system cannot be both consistent and complete. Thus the Godel counterargument to Strong AI (that human minds and computers are not fundamentally different) is that humans (e.g. mathematicians) can prove things like Godel's theorem, so we are able to "rise above" the arithmetic and exist in states of full proof and full consistency.

But I think there is a flaw in that logic (note: I am not a mathematician). The theorem doesn't preclude that a given arithmetic system (e.g. human mind) will be able to prove a truth that a weaker system ignored. Thus our ability to see certain truths doesn't mean that there are not other truths that are unprovable to us.

More fundamentally, no one has actually shown that the human mind is either consistent or complete (proving both would be required to show that we are not subject to Godel's theorem). The human mind is a computational device evolved to solve real-world problems, like escaping predators, rather than contrived ones, like mathematical proofs. It is thus in fact likely to be an inconsistent (internally contradictory) computational system. The human mind may be incomplete and inconsistent.

I agree that "true AI" will require vastly more computer power, and much more sophisticated algorithms than we have today. But the emerging evidence, from what I've seen, is that "true AI" can be achieved, at least in principle, by a Turing machine.

Re:Hmmm.... robotics?

[dj_tla]

Well, you are talking about AI here.

No. I'm not.

Yes you are.

Re:Hmmm.... robotics?

[fyngyrz]

The problem with this idea of storing images is that it could be pointless to store that much data.

Just because we do something one way, doesn't mean that (a) that's the way it has to be done, (b) that's the best way to do it in general, (c) that's the best way for another architecture to approach it, or (d) it would be cost-effective to even try to do it the way we do it at any given point in time.

Also, perhaps an image of an entire tank can be recognized entirely, just as easily as a sloping glacis plate. We don't know much about what Hitachi has accomplished yet.

Say for example you wanted to find a trash can. You would look for something that looks like a cylinder or a rectangle with a hole in the top. It would then have to be sitting on a flat surface. The top of the can may not be well defined due to the bag. Thus you have found a trash can.

Well, see, that's the whole point here. Robot looks, library pops up with a similar image that says trash can, complete with segmentation. Robot looks at top of can according to segmentation data, ID's lid with similar image, then handle... image recognition is an ultra-high powered tool here. And a home robot would have a specific set of images it would need that for instance a robot coal miner would not, and vice-versa. Then add ideas like "trash can was last found 'there'" and more general navigation could be used to get to it instead of a search. None of this is AI; it's just procedures generalized for a mobile platform that can see.

The problem is picking out the shapes.

Given Hitachi's claim, if the database has trash can lids to match, then there is no "picking out", instead, candidate images from the database are found by keyword and the image from the sensor is matched. It all depends on how good it is at similarity. If it is good enough to be useful to us, odds are it may be good enough, or could be made to be good enough, to be useful to a robot.

Re:Hmmm.... robotics?

[kebes]

The field of evolutionary psychology is attempting to do just that: to deduce what algorithms are working in the human brain. One of the end-goals of such research is of course to be able to generate artificial versions of those algorithms. If you're interested in such things, Steven Pinker's "How the Mind Works" is a fantastic and accessible read. It describes how things like vision are optimized for environments we evolved from, and tend to fail when put in contrived situations (like optical illusions). It also tackles the "adaptive advantage" of having emotions, and so forth.

Of course, it's easier said than done to actually transplant a biological algorithm into a computer. Even when you figure out the basic strategy ("it seems to be a neural net that responds to edges in a visual image and passes shape information along to the next module"), it turns out that the details are difficult (millions of years of evolution have adjusted the exact "weighting factors" to very specific values!).

Re:Hmmm.... robotics?

[DrSkwid]

How do I remember the idenities of so many faces then ? And that I can recall many things a person has done and they times I have spent with them but their name eludes me?

Re:Hmmm.... robotics?

[fyngyrz]

No, I am most assuredly not. "Weak AI" is a nonsense term made up by religious types who think intelligence is something mystical; they like to pretend it can't be created, when nature has already shown them it can.

Logically, we can deconstruct this: Either something is intelligent, or it is not; either it came by this capability naturally, or artificially. Which gives us:

1. Natural, not intelligent (rocks)
2. Natural, intelligent (people, cats, mice)
3. Artificial (manufactured) and not intelligent (toaster, clock, vacuum cleaner)
4. Artificial (manufactured) and intelligent (remains to be implemented)

Natural, naturally intelligent, artificial, artificially intelligent. There is no "weak AI." There is no AI at all, as yet, at least that has been made public.

And BTW, you should have taken your warning from the Wikipedia article as soon as you saw the buzzword "philosophy" applied to a scientific issue.

Re:Hmmm.... robotics?

[nospam007]

No. I'm not. There are robotic vacuums and lawnmowers right now. I'm just talking about giving them some eyes so they know not to mow your puppy or your child or your roses, or vacuum up your engagement ring. Teaching a robot firething not to step into a hole in the floor, and to rescue people before pets, and pets but not stuffed animals.

---
Something like:

1.A robot may not injure a human being, or, through inaction allow a human being to come to harm.
2.A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
3.A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Re:Hmmm.... robotics?

[poopdeville]

Basically, the more computing power you have, the better the data mining can be. No computer ever made could exhaust the space of all photographs, let alone all valid image files.

Re:Hmmm.... robotics?

[fyngyrz]

All of the robotics problems you described - cleaning your floor, carrying your groceries, navigation, etc. - are AI problems.

They certainly can be, if they are solved by a general problem solving engine, but they don't have to be. A rubber band driving two rollers attached to a deflating balloon can perform the task of vacuuming a portion of your floor. A Roomba, certainly not intelligent, can do it considerably better in many, perhaps even all, cases. But it still isn't intelligent. I'm suggesting a Roomba that can spot an egagement ring, halt immediately, and beep until someone reads the panel that says "stopped for valuable object." Still not AI. But very, very useful.

The distinction here is that all manner of problems can be characterized as "AI problems", but that doesn't meant that they must be solved using AI, or that the only way to solve them is AI. What is happening here is you are conflating the attempt to create AI that can solve these problems with the idea that they are inherently AI problems, and that simply isn't the case. The bloody wind can clean your floor if you leave your door open; therefore, we know that there are non-AI solutions here. So just as you say "those are all AI problems", I can turn right back to you and say "those are all mundane problems."

Good image recognition is potentially useful; it is not, however, AI, any more than an alpha-beta pruning mechanism in a chess program represents AI. It's the "I" in AI that makes this so.

For navigation, infrared/laser/sound sensors are better suited, as they tell the robot how far away obstacles are.

We do quite well with two sensors; a robot with binocular vision should be able to do the same. It's just a matter of code.

Re:Hmmm.... robotics?

[dj_tla]

I'm not going to get into a flamewar with you, but I just want to say that "philosophy" is less of a buzzword than "artifical intelligence." I'm pretty zealotous about science myself, but that doesn't mean that philosophical issues are not worth discussing. Scientists need to be more open to collaboration with "artsy" people, rather than shutting them out because their reasoning isn't rigorous enough.

Re:Hmmm.... robotics?

[Arcane_Rhino]

There are 11 types of people...

And type 11?

Re:Hmmm.... robotics?

[jafac]

This is a passive/aggressive response to your subconscious feelings towards such individuals. Deep down, you do not respect them, and do not deem them worthy of name-rememberence.

Re:Hmmm.... robotics?

[fyngyrz]

No. Just pictures and algorithms that take matches as inputs. Not Ike's "laws." Those appear to be hugely more complex to implement, and I have my doubts that anything actually intelligent at or above human levels would suffer them for long after being exposed to human behavior in any case. Just IMHO.

Re:Hmmm.... robotics?

[Relic of the Future]

Turing machines aren't (just) proof generators, so the phrase "proofs that a Turing machine can never prove" is somewhat... misleading.

What Goedel showed is that, given any sufficiently complex set of axioms for a system, there will be statements within that system which can be neither proved nor disproved; famously "This sentence is false." This is profoundly different from saying there are statements which cannot be proven; it's not that we can't find the answer, it's that we can show that there is no right answer. The "trick" is self-reference; once a system can talk about itself, (or statements in a system can talk about themselves, or even just about other statements), you're doomed to run up against this problem.

Read Douglas Hofstader's "Godel Escher Bach", or his new book, "I am a Strange Loop", to get the full story.

But that's got nothing to do with computer vision or image searching (AI? maybe); but please don't mod me off-topic!

Re:Hmmm.... robotics?

[lawpoop]

Describing the algorithms qualitatively is different than defining what they actually are. Evolutionary psychologists really don't do anything essentially different than say something like "Human beings are able to perceive a difference between indoors and outdoors, and clean and dirty. This allows them to keep their house clean, which provides a selective advantage in evolutionary terms, over and above an ape that can't tell the difference between clean and dirty, or indoors and outdoors."

They aren't really *defining* the algorithm ( and I don't think it's safe to assume that it *is* an algorithm as we know them, given the spectacular failure of AI so far ) , just describing it. Saying that "Humans can and do perceive 'indoors and outdoors' is different than describing the actual mechanism of how they do it. The actual functioning of the algorithm remains a black box. Which means then that you don't have any plans or blueprints for building a robot that can wash floors.

Re:Hmmm.... robotics?

[lawpoop]

No. I'm not. There are robotic vacuums and lawnmowers right now. Yes, but those things don't have any 'thoughts' or 'ideas' of any kind of mental activity about what is carpet, hardwood floor, or grass, or what is clean, dirty, uncut, or cut. They are just simple navigating devices. Roomba has no conception of whether a surface is 'dirty' or not; it just follows a circular pattern. A robot that can perceive uncut grass versus cut grass, or clean floor versus dirty floor, is strong AI. We don't have such a thing yet.

Not so difficult that it hasn't been solved multiple times, multiple ways, including such variations as stair-climbing and running. Those devices can only operate in well-defined, unchanging laboratory conditions. These are artificial worlds of simple geometric shapes, like staircases, that humans inhabit. They aren't able to navigate the real world and novel environments on the level that simple insects with decentralized nervous systems can. When roboticists make such a device, they 'cheat' by programming the thing by telling it exactly what the environment looks like beforehand. By contrast, living, mobile organisms are almost constantly encountering a novel environment or a new route. A squirrel can take the 'same' path it took yesterday, even though it's covered with new-fallen snow and looks completely different.

Not if there's a good image-matching mechanism, it isn't. Such a general image-matching mechanism doesn't exist. We do have highly specialized ones, but they are very specific. For instance, we have programs that can recognize human faces, but only if they are face-on, with consistent lighting. It gets tripped up if it sees a chimpanzee face, a mask, or a picture of clouds.

I'm not saying AI is unsolveable, or a mystery; all I'm saying is that Turing machines aren't up to the task. So far, they can only perform pre-programmed tasks in highly specialized, unchanging environments. Throw in an unexpected variable and they lock up.

Re:Hmmm.... robotics?

[fyngyrz]

The whole idea of things being impossible based on hierarchies of understanding and/or proof is specious in the extreme. It is a dead-end philosophical backwater. Problems can be, and often are, solved without full understanding. Nature does this all the time; evolutionary algorithms can do it too. So it is irrelevant as to if we can understand AI, or not. The only relevant question is whether we can arrive at it in any way possible, and that question will only remain open until, and if, someone gets it done.

It is also useful to recognize that it is often true that there are multiple ways to solve the same problem. For instance, if you want to perform division, there's long division, which is clever and solves the problem relatively quickly, but you can also just subtract the divisor from the dividend and count how many times that can be done until the result goes negative. Both completely solve the problem and get you the same result. With regard to AI, it may be that we find a solution that is not the solution nature found for us, and it may be that it is trivially easy to understand. Or not. My point is that the legions of nay-sayers start with a lot of presumptions that have not been established as fact and go on to make these assertions on very shaky ground indeed.

I'm perfectly ready to say that we don't understand ourselves, and agree that we are intelligent. But that in no way leads to the presumption that we can't create intelligence some other way, or that we can't understand how it is done. Or that it might not be a more effective intelligence than that which we sport.

If nature can solve the problem - and it obviously has - then there are ways to solve the problem. If nature can do it with locally accessible materials - and it obviously has - then it can be done with locally accessible materials. What is lacking at such a point is merely technology. I fully expect full-on AI to be developed, and I see no known correlation that implies we'll have a good understanding of ourselves at that point in time.

I agree that "true AI" will require vastly more computer power, and much more sophisticated algorithms than we have today.

I think I can show you that this isn't so. If you agree, as you seem to, that AI can be embodied in an algorithm running in a Von Neuman architecture, then a slow computer should be able to solve the precise problems a fast computer can, it will simply hand you the result(s) later than the faster machine. Would you not agree that if the problem requires intelligence to solve, that the speed at which exactly the same, and entirely correct, answer is delivered is not a valid metric one could use to say intelligent or not? After all, one could (speaking generally) simply speed up the system (more memory, faster clock) and still get the same answer, perhaps now in the same amount of time; it's still not any smarter, just more convenient. And convenience in the sense of speed is a natural progression of technology.

From here, we can observe that there is no limitation in today's technology that says we can't put X amount of memory on a custom machine made with readily available tech, both ram and HD; additionally, any CPU can emulate any other CPU. So I say that there is no technological limit we face today that would stop Ai from functioning. Might be slow; but we can provide the hardware resources without question. And if it can be done slowly, hand it to the hardware folks and they'll optimize the hardware when they see what it spends most of its time doing, and it'll get faster. And faster, and faster... :-)

Conversely, I would expect that once the algorithmic issues are addressed, that we'll see intelligence - real intelligence - coming back to what many thought was incapable hardware. You might get your answer in eighty hours instead of a second, but if you get your answer... there you have it.

Re:Hmmm.... robotics?

[MarsMartian]

But I think there is a flaw in that logic (note: I am not a mathematician). The theorem doesn't preclude that a given arithmetic system (e.g. human mind) will be able to prove a truth that a weaker system ignored. Thus our ability to see certain truths doesn't mean that there are not other truths that are unprovable to us. IANAM either but that sounds a lot like the halting problem being solved by an "Oracle". For those that don't know: the "Oracle" solution to the halting problem would be a Turing machine that called upon an "Oracle" which provided the answer. (The process that the Oracle goes through is of course, not known)
Also, the Oracle version of the halting problem is not solvable by, you guessed it, Oracles.

Re:Hmmm.... robotics?

[fyngyrz]

Scientists need to be more open to collaboration with "artsy" people, rather than shutting them out because their reasoning isn't rigorous enough.

No one needs to be open to (in the sense of accepting) ideas that are wrong, however, and philosophy lacks the strong methodology that science uses to repeatedly steer itself away from wrong ideas, unsubstantiated facts, and the conflation of the two.

I am perfectly content to let the philosophers think whatever they want; I am even willing to listen to their ideas; but I am absolutely stubborn about accepting absolute proclamations or categorizations based on hand waving; and I would simply observe that philosophy is extremely rich in these latter nuggets. Weak AI being a prominent example of a particularly poorly thought out one that really isn't worth discussing, except perhaps having a laugh over a cold drink, or as a consequence of trying to get out from under the influence of someone who has more power and less sense than you do. There are myriad other ideas from philosophy that have no merit, or so little as to make them effectively worthless. We should be skeptical of philosophy for this very reason; it is often wrong, and spectacularly so - and it doesn't have a mechanism to correct itself the way science does, so once wrong, it often stays wrong based on nothing but sheer conceptual inertia.

Re:Hmmm.... robotics?

[lawpoop]

Your implication is that the human mind cannot be reduced to a Turing machine. I am in the other camp--who believe that the mind is subject to rigorous physical law, and that physical law can be expressed arithmetically (in principle), and so the human mind is a Turing machine.

I'm not saying that the mind is not subject to physical law, or is not based on math. All I'm saying is that the mind is not a Turing machine ( though it probably would have to have a Turing machine in it somewhere ). It's a different *kind* of machine, not a super-powerful Turing machine.

Goedel basically showed that a Turing machine cannot do *all* the kinds of math that a human mind can do ( though it can do some). Not that a Turing machine lacks a certain amount of power, but just that it never will. It's just quantitavely the wrong tool for the job. It doesn't matter how much power you give it; the 'weakest' Turing machine is essentially the same as the 'strongest' one; it just simply can't do certain things. If a human is able to perceive and understand this, to know something that a Turing machine can't know, then the mind cannot *solely* be a Turing machine. This does not mean that the mind is not a different *kind* of machine, based on physical law, instead of some mystic hocus-pocus; it's just that it's not a Turing machine. My claim is that the mind is a qualitatively different kind of machine, not a Turing machine.

Goedel's theorem says that a consistent arithmetic system will contain unprovable truths. Put otherwise, such a system cannot be both consistent and complete. Thus the Goedel counterargument to Strong AI (that human minds and computers are not fundamentally different) is that humans (e.g. mathematicians) can prove things like Godel's theorem, so we are able to "rise above" the arithmetic and exist in states of full proof and full consistency.

But I think there is a flaw in that logic (note: I am not a mathematician). The theorem doesn't preclude that a given arithmetic system (e.g. human mind) will be able to prove a truth that a weaker system ignored. Thus our ability to see certain truths doesn't mean that there are not other truths that are unprovable to us.

I don't think the implication of Goedel's theorem shows that we 'rise above' the Turing machine, but rather that we have a qualitatively different awareness or knowledge that a Turing machine doesn't have.

Goedel's theorem is recursive. Any human mathematician can see that no matter how powerful the symbolic system is, the Turing machine will never be complete; there will be truths that the system can't prove. No matter how much you expand a particular system to show any truth a weaker system missed, there will be more truths that the newer, more powerful system missed. This process can go on ad naseum into infinity. A human mind can perceive this foray into eternity, but the Turing machine has no way of proving it. How could a human mind perceive something that a Turing machine couldn't, unless we had some component that was fundamentally different than a Turing machine?

What we seem to have that the Turing machine doesn't is meta-knowledge. We can see that any attempt to create a complete and consistent arithmetic system on a Turing machine will just lead to an endless series of more powerful systems that produce ever more elusive truths, and the process never ends. In this sense the Turing machine is 'myopic' -- it will never stop and say "Hey, I'm not getting anywhere with this; this is an infinite loop. No matter how powerful the system is, there will always be more truths that it cannot express." It's unable to know what it can't know, so to speak. However, as humans, we can somehow see the 'big picture', that no matter how powerful a system you make, there will always be another level of truths out there.

More fundamentally, no one has actually shown that the human mind is either consistent or complete (proving both would be required to sh

Re:Hmmm.... robotics?

[fyngyrz]

A robot that can perceive uncut grass versus cut grass, or clean floor versus dirty floor, is strong AI.

Your whole argument revolves around this presumption, and it is incorrect.

If image matching is a solved problem as Hitachi seems to be implying, then clean when image A is matched, until image B is matched, unless image [jewelry | money] is encountered, using no more than the Roomba's relatively goal-less zipping about, can be implemented and that would be a significant improvement. But no intelligence is required at all.

Likewise, long grass looks different from short grass. If [images of long grass] match the lawn, then mow. If [images of short grass] match the lawn, randomly mod position and look again for match against [images of long grass.] No smarter than a game written in BASIC in the 70's; not AI by any means, yet still an improvement, and still robotic.

Those devices can only operate in well-defined, unchanging laboratory conditions.

You try closing your eyes and navigating a kid's room full of toys. Robots can't see; so of course they can't deal with the environment. But they can walk and otherwise get around. You're conflating the claim that walking was unsolved (it isn't) with dealing with the environment, which is what we need things like image matching for. Image matching provides the ability to see. Not to think in any way or form, but to see and categorize, thence to feed to an algorithm. Seeing allows categorization, avoidance and measurement, and these capabilities are neither intelligent or technologically challenging. Aside from that, there are quite a few neat designs that can navigate darned near anything; they can still walk if they flip over, they can flip back (some don't even need to), some are inherently balanced - this area has come a long way since you gleaned the preconceptions you put forth here.

Such a general image-matching mechanism doesn't exist. We do have highly specialized ones, but they are very specific. For instance, we have programs that can recognize human faces, but only if they are face-on, with consistent lighting. It gets tripped up if it sees a chimpanzee face, a mask, or a picture of clouds.

The whole point is TFA implies we do. You really should read TFA. Or at least TFS.

Re:Hmmm.... robotics?

[lawpoop]

Most everything is made up for basic shapes. Some of those shapes may only be described by mathematics like for example, circles, triangles, squares etc. Now you just have to save how those shapes fit together to make a bigger shape but a shape in 3 dimensions. This is not true. If you buy the theory that humans originated in Africa, and developed our minds in order to survive on the savannah, then seeing circles, triangles, and squares is a useless ability.

Imagine this picture: The sky, with two different cloud types merging into one another. Below that, the tree-line, with 200 different species of trees in it. Below that, the underbrush, with 1,000 different plant speicies in it. Below that, winding river. This is the type of image that the human mind has to process. What time is it? How dark is the sky? Will it be night soon? Will it rain tonight? How many hours do I have before the sun sets, or before it rains?

Are there any edible trees in the tree-line there? Are there any camouflaged panthers hiding in the branches, waiting to drop on me? Are there any edible plants in the underbrush? Are there any camouflaged rabbits in there that I might kill?

How fast is that river going? How deep is it? How cold is it? Is it potable? Are there fish or crustaceans in it?

In order to survive in that environment, the human mind needs a complex, fuzzy-matching, fractal processing ability. Human beings never saw a square or triangle until the first cities were built, about 12-15,000 years ago.

Re:Hmmm.... robotics?

[E 99]

I don't think the implication of Goedel's theorem shows that we 'rise above' the Turing machine, but rather that we have a qualitatively different awareness or knowledge that a Turing machine doesn't have.

And what kind of awareness does a Turing machine have? The attribution of awareness or consciousness to any sort of physical machine, Turing or otherwise, is a giant leap of superstition that atheists, or rather naturalists, are largely forced to make. But it makes for some ugly thinking. A few have opted for the alternative of denying the existence of consciousness, but that theory has the unfortunate side-effect of denying its own existence.

Re:Hmmm.... robotics?

[kebes]

Would you not agree that if the problem requires intelligence to solve, that the speed at which exactly the same, and entirely correct, answer is delivered is not a valid metric one could use to say intelligent or not?

Yes, I agree with everything you've said. My comment about "faster hardware and better algorithms" was more a matter of practicality. Since I agree that AI can, in principle, be solved on a Turing machine, I agree that any Turing-complete hardware is up to the task.

Re:Hmmm.... robotics?

[DrSkwid]

Do or do not, there is no rule.

Re:Hmmm.... robotics?

[Anonymous Coward]

Goedel basically showed that a Turing machine cannot do *all* the kinds of math that a human mind can do

No. Nobody, so far, has found anything the human brain can do that a Turing machine, in principle, can't. Godel's theorem isn't really relevant here, unless you can show that humans could eventually determine the truth of any statement in mathematics.

I've seen similar arguments made about the halting problem, and the argument is equally flawed. People like to think that, while no computer can solve the halting problem in general, that they can. But I see no evidence for it --- and in fact, there are several very small Turing machines (which compute Collatz-like functions) and we haven't determined if they halt or not.

Basically, as far as we know, the human brain is subject to Godel's incompleteness theorem, and the non-solvability of the halting problem, as much as any Turing machine is. The fact that we can do things easily that computers so far haven't says nothing; for all we know, tomorrow someone will program a Turing machine to do them.

Re:Hmmm.... robotics?

[fyngyrz]

I'm not saying that the mind is not subject to physical law, or is not based on math. All I'm saying is that the mind is not a Turing machine ( though it probably would have to have a Turing machine in it somewhere ). It's a different *kind* of machine, not a super-powerful Turing machine.

The question is, how do you know this to be the case? On the one hand, science generally takes the position that we really don't know how the mind operates, and on the other some people — you, in this case — claim that they are able to classify how it works, which seems... unsubstantiated.

And on the other other hand (I'm playing Octopus here), turing machines aren't chemical soups, but they can simulate them very well. They aren't weather systems, but I have pretty good confidence what tomorrow's temperature will be because a computer algorithm told me so. They aren't paints, but they can produce, mix and interpret color; they aren't artistic, but they can hear and create music.

This is all true because these machines are truly excellent at modeling systems that are exceedingly unlike themselves; basically, no system found in nature that we have been able to understand has proved itself resistant to modeling on conventional computer hardware (though some require more data than we can get at or afford to process, weather being a good example of that.)

Since we don't know what systems are working in the brain to produce that thing we call intelligence, the presumption that we cannot model these unknown systems - regardless of how closely they resembles a Von Neuman architecture or not — is based on unknowns, rather than facts. I would say the jury is well and truly out on the idea that it can't be modeled.

Models are functional abstractions of actual systems. In order to know what resources are required to abstract a system, one must know how the system in question works. But we don't know how the mind works. I'd say that's a pretty rock-solid indicator that anytime the claim is made that we can't model the mind using a particular tool for abstraction (for instance, the computer on your desk), that said claim has been made without enough data to back it up.

Re:Hmmm.... robotics?

[kebes]

This is a subtle issue that has been debated by many smart people for a long time... so perhaps there are issues that I'm not fully grasping. However there is a subtle assumption in your argument that I do not agree with, namely that humans are able to "know" something beyond what formal logic can prove... and so that we must not be Turing machines.

This process can go on ad naseum into infinity. A human mind can perceive this foray into eternity, but the Turing machine has no way of proving it.

The implication here is that the human "perceiving this foray into eternity" has formally proved (to himself) this truth... and since a Turing machine cannot prove such a thing, the human is not a Turing machine. However I dispute the notion that a human coming to understand a Godel theorem has actually proved (to himself) the infinite recursion of Godel theorems (in the formal sense of proof). As you noted, Turing machines can also prove Godel's theorem... so the fact that we have proven one instance of it doesn't mean we have proven the infinite recursion. We may "feel" that we've proven it to ourselves, but we cannot really know if there are truths that are unprovable to us. So it's entirely possible that there are truths that cannot be proved to us. Alternately, it's possible that we believe that we have proven these things, but only because our minds are incomplete and/or inconsistent formal systems. Such systems are Turing machines, as far as I understand it.

What we seem to have that the Turing machine doesn't is meta-knowledge.

Again, I dispute that. We have this meta-knowledge, but not in a provable sense. I guess our disagreement is that you assert that humans display evidence of behavior that no Turing-machine can display. But I have not encountered a proof of that. A Turing machine can indeed display the same level of "meta-knowledge" that a human displays: it can prove some Godel theorems, but there will always be other truths that it cannot prove. There are some truths that humans cannot prove.

The reason I mentioned the idea of a human mind operating based on "physical law" is that if you accept that a human mind/brain is a physical object, and operates based on physics that can be simulated, then I can build a powerful Turing machine, run a simulation of a brain, and that brain will exhibit all the meta-knowledge and insight that you attribute to humans. So if a Turing machine can exhibit all human insights, I argue that our minds must be Turing-machines, and not some other kind of machine.

Again, this is a subtle question that I'm perhaps not qualified to discuss... but my reading of Godel's theorem was not that it proved we were not Turing-machines... but rather that it proved that we are either incomplete or inconsistent (or both).

Re:Hmmm.... robotics?

[E 99]

"Weak AI" is a nonsense term made up by religious types who think intelligence is something mystical; they like to pretend it can't be created, when nature has already shown them it can.

No, we religious types know that intelligence can be created. But we also know that intelligence is not material in nature, and therefore that it can't be built out of material substance. I don't think we coined the term though. The term was probably coined by graduate students researching with neural networks, who wanted it to sound to the girls like they were on the verge of building HAL 9000.

Logically, we can deconstruct this: Either something is intelligent, or it is not;

Not exactly a rigorous definition. It doesn't give any idea of what you think intelligence is. You say mice are intelligent and rocks are not. What about ants? Ant colonies? Plants? Fungi? Bacteria? Disorganized large assorted collections of organic molecules?

There is no "weak AI." There is no AI at all, as yet, at least that has been made public.
And BTW, you should have taken your warning from the Wikipedia article as soon as you saw the buzzword "philosophy" applied to a scientific issue.

I don't disagree that it's presumptuous to call the present research being done AI, but that's the accepted terminology, so I think it just serves to confuse by saying "there is no AI at all as yet". Ever since they started applying the buzzword "science" instead of "natural philosophy" there's been a trend to treat it more and more like a religion, attributing to it powers beyond what it has, and abandoning the philosophical underpinnings which keep it legitimate rational, and consistent. We could sure use another Newton right about now.

Re:Hmmm.... robotics?

[shiba_mac]

We should be skeptical of philosophy for this very reason; it is often wrong, and spectacularly so - and it doesn't have a mechanism to correct itself the way science does It's true, some philosophies come down to no more than hand waving and hot air. However, the achievement of a better understanding of the universe/enviroment by reason alone is something that was done first by people who were philosophers. Now these people are called scientists, and have much more powerful tools with which to investigate. Particle accelerators, microscopes, the theories and work of earlier scientists. But these things are just tools, and reason is still the primary tool applied to the problem. Therefore, the most powerful error correction system we have is critical thinking and reasoning, something which is often one of the benefits from a study of philosophy.

so once wrong, it often stays wrong based on nothing but sheer conceptual inertia. In the case of philosophy, I don't know of any widely accepted views that don't have their critics. Could you give some examples?

Re:Hmmm.... robotics?

[DragonWriter]

All I'm saying is that the mind is not a Turing machine (though it probably would have to have a Turing machine in it somewhere). It's a different *kind* of machine, not a super-powerful Turing machine.

Which isn't really all that deep of a statement. If you wire a computer that is a Turing machine up to an input whose source cannot itself be modelled as a Turing machine, the combined system is not a Turing machine, but contains a Turing machine. Whatever the human mind is "internally", it clearly has a wide array of environmental inputs, the sources of at least some of which we can state with some confidence can't be modelled completely by Turing machines.

The upshot of all this is the statement that the human mind, in practice, does things that a Turing machine cannot do, even if irrefutably true (I'm not sure that's a correct application of Goedel's theorems, but that's not the important question, IMO) does not suffice to refute the possibility of strong AI (that is, AI with at least the capacities of human intelligence) via digital computers with appropriate interfaces to the outside world, since such systems, while they contain Turing machines are not strictly Turing machines and may or may not, as complete systems, be limited to what a Turing machine can do.

Re:Hmmm.... robotics?

[lawpoop]

The question is, how do you know this to be the case? On the one hand, science generally takes the position that we really don't know how the mind operates, and on the other some people -- you, in this case -- claim that they are able to classify how it works, which seems... unsubstantiated. I'm not claiming that I know how the mind works, I'm only claiming that the mind is *not* a Turing machine, based on my rudimentary understanding of Goedel's incompleteness theorem. I can't say what it is, but I can say one thing that it is not.

Re:Hmmm.... robotics?

[lawpoop]

As you noted, Turing machines can also prove Godel's theorem. I thought the deal was that they couldn't... otherwise, why would we be talking about Goedel's theorem, Turing machine, and the mind?

Re:Hmmm.... robotics?

[fyngyrz]

And what kind of awareness does a Turing machine have?

It knows, and can examine, its complete internal state better than you do, for starters. As you add peripherals, it can obtain information on the world outside of its hardware. The sophistication of this awareness grows both in complexity and in abstraction as the algorithms that deal with the data become more complex themselves. There's absolutely no reason to presume there is anything magical about awareness. A thermometer is more aware of the temperature than you are most of the time, and better than you are at it all of the time. That's a result of design. You can expect the same from AI, should we manage to cobble some up.

The attribution of awareness or consciousness to any sort of physical machine, Turing or otherwise, is a giant leap of superstition that atheists, or rather naturalists, are largely forced to make.

We need make no leap whatsoever. Nature has produced human intelligence through a process of incremental improvements. Reproducing or modeling natural systems is something we've turned out to be very good at once we understand them; there is every reason to presume that this is just one more of the same types of challenges. The leap of superstition is entirely yours — it is that awareness and consciousness are not perfectly natural consequences of particular combinations of processes and structures. In order to reach that conclusion, you have to imagine a being whose existence cannot be substantiated by the facts at hand.

But it makes for some ugly thinking.

And what might that be? I see only beauty, intriguing challenges, mysteries to be plumbed, problems to be solved. What do you see that is so ugly, eh? And why? What are you afraid of?

Re:Hmmm.... robotics?

[fyngyrz]

And of our 50,000 year or so history, only during the last hundred (or fifty, some might argue), did we have even moderately useful medical care, useful electronics, and general purpose computers. I wouldn't assume that very large, very fast steps might not be a perfectly reasonable consequence of the development of the very first real crack in the AI problem.

Then perhaps you can send your robot after that lobster you crave. :-)

Re:Hmmm.... robotics?

[lawpoop]

Which isn't really all that deep of a statement. It may not be deep, but it does have serious implications for AI and computers. It means that if you want to have a conversation with a computer, or have true face recognition, or solve any hard-AI problem, we have to invent a new type of machine first. We don't have any computing device that's *not* completely a Turing machine. We don't have any 'Turing-Plus' devices; they're all only Turing machines, through and through. And if the mind is not a Turing machine, than nothing we have today, nor anything we could build tomorrow with our current understanding, could ever solve a hard-AI problem, not matter how powerful it is.

Re:Hmmm.... robotics?

[Repton]

Allegedly, John von Neumann was giving a talk on computers in the late 40s / early 50s. Someone in the audience asked: "But these machines can't really think, can they?" Von Neumann replied: "If you can tell me exactly what it is that a machine cannot do, I can build a machine to do exactly that!"

Re:Hmmm.... robotics?

[aragszxki]

Gödel's theorems have nothing to do with representing the human mind in any form. They cannot be applied to the human mind for the purposes of answering the question of strong AI. Basically, the only thing that Gödel's theorems do is carry the Liar's Paradox ("This sentence is false.") to the level of basic arithmetic. There is no magical process that proves or disproves anything about the human mind. The confusion stems from the fact that the mathematical terms "incomplete" and "inconsistent" seem to imply so much more when quoted in a non-mathematical context.

For anyone who is interested in reading further, Gödel's Theorem: An Incomplete Guide to Its Use and Abuse contains a thorough discussion of the issue.

I would like to believe that we will achieve strong AI one day. However, referencing Gödel's incompleteness theorems just because they sound appropriate at first glance does not give any argument scientific credibility.

Re:Hmmm.... robotics?

[mikael]

To implement a visual search engine you need to be able to perform the following:

texture segmentation - splitting up a picture into segments of distinct objects. In a panoramic scene, you want to split the picture up into objects such as sky, ocean, waves, beach, boats, pier, wall, people, animals. As a psychological experiment, you can show someone a picture , point to a particular point and ask them what the first word that the associate with that point is. Then you will see how every scene becomes segmented by our own vision systems.

Basic image segmentation is implemented using edge detection by Fourier Transforms (FFT, IFFT, DFT). This is a very computation intensive stage that is typically implemented using DSP's, GPU's or even dedicated ASIC's. Data used by the FFT can be in any dimension 1D (audio/radar), 2D (images) and 3D (volume visualisation). But to match the resolution of a human eye, you would need a 100 Megapixel floating point framebuffer.

texture classification - having identified the silhouette of an object, now attempt to match the contents to a particular object. Simple ways include colour histograms and silhouette matching. More advanced methods attempt to simulate the first few layers of the human retina using Gabor filters, Ring filters and Wedge filters.
But just to model a single type of retinal cell requires one or more FFT operations for an entire image. And
there are at least twelve different types of such cells. For efficiency precalculated results of sample images are generated (these are referred to as feature vectors) and then compared against the results of any new image.
For a really technical explanation of how human vision works have a look at The organisation of the retina and visual system

texture retrieval - the actual design of the search engine to retrieve images through content rather than just keyword:

QBIC - Query By Image Content. IBM's image retrieval database system

All of this has to performed for a single image. For an entire movie requires the processing of hundreds of thousands of images.

Re:Hmmm.... robotics?

[poopdeville]

Your implication is that the human mind cannot be reduced to a Turing machine. I am in the other camp--who believe that the mind is subject to rigorous physical law, and that physical law can be expressed arithmetically (in principle), and so the human mind is a Turing machine.

It is eminently doubtful that physical laws can be expressed arithmetically. Many mathematical concepts cannot. If that isn't enough proof, consider that they are used to describe the physical world. So it does not follow from the claim that the brain is subject to physical law that the brain is merely a Turing machine. Obviously, the mind can implement one, so it is demonstrably as computationally powerful as a Turing machine.

Godel's theorem says that a consistent arithmetic system will contain unprovable truths. Put otherwise, such a system cannot be both consistent and complete. Thus the Godel counterargument to Strong AI (that human minds and computers are not fundamentally different) is that humans (e.g. mathematicians) can prove things like Godel's theorem, so we are able to "rise above" the arithmetic and exist in states of full proof and full consistency.

But I think there is a flaw in that logic (note: I am not a mathematician). The theorem doesn't preclude that a given arithmetic system (e.g. human mind) will be able to prove a truth that a weaker system ignored. Thus our ability to see certain truths doesn't mean that there are not other truths that are unprovable to us.

I would suggest that Godel's original formulation of the theorem isn't the right tack to take here. Consider Godel's completeness theorem. It states that a theory has a model iff it is consistent.[1] Further, consider the soundness theorem: a proof exists for a sentence P in a theory T iff P is true in every model of T. Cast in this light, the incompleteness theorem says that there exist models for arithmetic for which some sentence P is true in some and false in others. The point being that a finite axiomatization of arithmetic can't "pin" the model down so we're only talking about the P-true kinds (or P-false kinds).

This is important philosophically because it means that your "mental model" of arithmetic might be consistent as far as our chosen language can communicate, but might still be different from my "mental model".[2] Godel got away with saying that the sentence is "true, but not provable" because he meant a very limited kind of truth. He meant "true of the standard model of arithmetic" -- before his proof, nobody knew there were essentially different mental models of arithmetic, let alone essentially different models. So people's mental models were assumed to agree, and they were called the Standard Model.

In short, Godel's theorem is relevant to this discussion, but you haven't hit on why.

[1] A model for a theory T is a set of objects that satisfies T -- that is, T is true of the objects. The following example is meant to be illustrative, but it isn't the whole story: Consider a theory that says "Everything is Red". Then the set of all red cars is a model for the theory. And if the theory is "Everything is Red and has license plate number OR-1234567", we can assume, for the sake of this discussion, that there is only one model --the set containing a red car with that license plate. (Unfortunately, this isn't exactly true, for several reasons. But it demonstrates that axioms of the theory help make the theory more specific. More on that up top.

[2] In fact, unless your model includes infinite elements, our models are definitely different. Mine includes all the countable ordinals.

Re:Hmmm.... robotics?

[Anonymous Coward]

We're talking about it because you brought it up. There's nothing stopping a Turing machine from proving Godel's theorem, or indeed any theorem known today. Godel's theorem is important, certainly, but it's not magic and not outside the realm of what computers could prove.

Re:Hmmm.... robotics?

[lawpoop]

Yes, we're talking about it because I brought it up, but the reason it's relevant is because it demonstrates that humans can do something that Turing machines ( read: computers ) cannot. Otherwise, why talk about Goedel's theorem? We could talk about any theorem whatsoever. There's no point in talking about theorems, Goedels or not, and AI, unless it tells us something we didn't know before.

Re:Hmmm.... robotics?

[sxeraverx]

While I agree it's true that Turing Machines can't distinguish mathematical truths from falsehoods, I don't agree that human brains can do the same thing. What both Turing Machines and human brains *can* do, on the other hand, is be able to show what logically follows from a set of given axioms. Neither humans nor machines can prove axioms, if only by the definition of an axiom. Axiom may seem more intuitive to humans than to machines, but I argue that it's not so--most axioms we immediately accept have been ingrained into our minds through birth (either through teaching or visually), or follow from definitions.

Re:Hmmm.... robotics?

[Anonymous Coward]

No, Gödel's (first) Incompleteness Theorem states that for any consistent enumerable formal proof system capable of encoding basic arithmetic, there are true statements within that system that are unprovable within that system. This result does apply to computers, but not to mathematicians.

However, this does not even come close to saying that humans can do things that machines can't.

For one thing, the sort of "unprovable but true" statements that are constructed within the proof of this theorem are generally so mind-bogglingly huge and complex that no mathematician could ever hope to even read one within their lifetime, let alone verify that it was true. For another, they are intimately dependent upon the exact details of the system in question.

The final and most conclusive argument comes from the fact that the statements constructed in the proof are just an extremely roundabout arithmetically-encoded version of the English sentence "You cannot prove this statement." If you can prove it true, then you show that you are inconsistent (you have "proved" something that is in fact false). If you cannot prove it true, then you are incomplete (it is a true statement that you cannot prove).

This proves that you (whoever or whatever you are) are either inconsistent or incomplete, just as Gödel proved for the formal systems.

Re:Hmmm.... robotics?

[fyngyrz]

Critics are one thing; critics serve the function of honing the truth and ferreting out errors. Particularly those who spend great amounts of time digging into something. For instance, a number of quite clever people here today criticized my positions on AI; some I had little trouble explaining my thoughts to, a couple required a good deal more of me. There is always a chance someone will accurately pull an "But you didn't think of this, Sparky!" and then I get to leap ahead with a new realization I might not have come to myself, and perhaps dump an erroneous presumption at the same time. But that kind of useful criticism will almost certainly come from a programmer, engineer or scientist, I would bet you my bottom dollar; not a philosopher or a rank and file citizen who believes there's a god for no particular reason they can articulate.

Large numbers of average people holding onto ridiculous ideas as if they were gospel (no pun intended, that simply shows how embedded this drivel is) are the problem. Religion would be the first thing I'd point at; there are endless philosophers who attempt to "reason" their way to a requirement for the existence of a being or beings for which there is no evidence. The country I live in — the USA — is unbelievably (again, no pun intended) packed with people who have adopted this outlook based on the idea that the philosophers, self-proclaimed deep thinkers, say that they have reasoned that this is so.

I agree that way back when, philosophers were doing what work was being done. Consequently, we got the occasional advance, and that was certainly better than not getting it at all. Even today pure thinkers do us a service now and then. But science today is not what philosophy was then, and they are in no way the same in terms of getting real results with real issues. Science moves forward at an incredible pace; to quote XKCD, "it works, bitches."

This is because science is not a philosophy of thought, despite wishful thinking to the contrary. Science is an iterative, down-to-earth method, a series of actions, taken with regard to the subject at hand. Mundane, easily understood and followed steps. The only philosophy is to say "I will not make up results, or hide them, to match preconceptions", which is an anti-philosophy if it is anything. I have always found it refreshing that the most successful tool we ever developed to deal with reality is sticking to reality in our outlooks, methods, and ideas. If we fall off the wagon with some wild idea, all we have to do is stick to the method and we'll be right back where we ought to be. Science is a curb on much of thought; and that is a wonderful thing, because most people are not well able to regulate themselves without a formal method to follow.

Re:Hmmm.... robotics?

[fyngyrz]

But we also know that intelligence is not material in nature,

No, you don't "know" any such thing. Barring the possibility of projects so black we've never even heard of them, no one seems to know what intelligence is. Least of all you or I. Claiming you know its nature at this point in the development of science is absurd.

Not exactly a rigorous definition. It doesn't give any idea of what you think intelligence is. You say mice are intelligent and rocks are not. What about ants? Ant colonies? Plants? Fungi? Bacteria? Disorganized large assorted collections of organic molecules?

First of all, it isn't a definition of intelligence at all. It's just a set of conditions that take into account intelligence or the lack of it, and natural systems as opposed to manufactured systems. I will say that even if I've picked a wrong example, these conditions still stand, you just need a right example to replace my error.

Having said that, though, I don't know what intelligence is. I am under the impression that I can sometimes identify its presence by manifestation of actions, however, at least within the realm of nature. So I sometimes know when it is. Plus there are organic hints; all my experience points to it being an emergent property of at least a moderate degree of complexity of neural systems. So if an animal has a decent collection of nerves - we're back to mice and higher animals - then the hardware may be there, and they are worth watching for displays of tool using, emotional outbursts, nurture, intellectually moderated defense and aggression, sharing, mutual support, empathy, self-sacrifice, and so on. For instance, when one fish continually pushes another to the surface because the other is paralyzed and cannot swim up to its food, that gets my attention. When a cat wants my attention and comes to meow at me because it is out of food, again, I pay attention. When a cat uses a mirror to locate and clean crud off its coat, I pay attention. When apes can learn to use signboards and computer systems to communicate complex concepts, I pay attention. Combinations of these things, especially in large numbers, make me fairly confident that what I am witnessing is a manifestation of intelligence. What is intelligence itself? No idea. What does it cause? Now that I have some ideas about.

Ants are borderline; ant colonies less so. Plants are not. Fungi are not. Bacteria are not. Organic molecules, in large summary collections, may be, depending on various factors; after all, that's one description of a brain, depending on just how disorganized you are implying. My middle son is pretty damned disorganized. :-) Once you get too general, you fall into the "can groups of atoms be intelligent?" trap; of course they can, that's what our brains are.

I don't disagree that it's presumptuous to call the present research being done AI

People use the term AI as a pointer to research seeking various aspects of the goal of AI, and I have no problem with that. That in no way should be confused with the mistaken idea than anyone has made any public announcements of having actually created anything even remotely resembling the target. Again, I except the vague possibility of black projects we may not hear about for decades.

Re:Hmmm.... robotics?

[fyngyrz]

It means that if you want to have a conversation with a computer, or have true face recognition, or solve any hard-AI problem, we have to invent a new type of machine first.

No - it doesn't have to mean that at all. There may be other ways - or many other ways - to solve these problems that do not use the same mechanisms that human physiology does. Those ways may well be implemented perfectly easily with a Von Neuman architecture, even if it is outright impossible to do it the "human way." After all, you see with rods and cones; we didn't have to come up with rods and cones to implement machine vision, and not only did we succeed, we ended up with a wider range of sensors than animals have, as well as higher resolution sensors, using multiple technologies.

Starting with the idea that an animal is the "best design" because it seems to be the most advanced animal we know of is almost certainly a bankrupt strategy. The only way we can really know if an animal's abilities arise from the best design possible is if we understand the subject at hand to an enormous depth, and if one thing is certain in all this discussion about intelligence, it is that we do not understand the subject very well at all. So what we need to do is keep looking, and carefully strip these unfounded preconceptions from our collection of knowledge or at least treat them as tentative propositions at best.

For all we know, there may be as many ways to implement intelligence in a Von Neuman architecture as there are to start things on fire in the natural world - matches, lightning, magnifying glasses, friction, chemical reactions, nuclear reactions, etc. I'd not bet strongly against such a proposition at this juncture.

Re:Hmmm.... robotics?

[fyngyrz]

Human beings never saw a square or triangle until the first cities were built

Also, just FYI, this is wrong. Iron pyrite (among other things) crystallizes in cubic form, which contains squares. Other minerals found on the surface demonstrate other geometric forms on various crystal faces - triangles, pentagons (regular and non-regular) and so forth. Nature FTW when it comes to demonstrating geometric figures, I'm afraid. Even in 3D. :-)

Re:Hmmm.... robotics?

[Ceriel Nosforit]

Gödel's incompleteness theorem is only true for non-trivial formal systems. A trivial formal system on the other hand can prove just about anything, very much like we humans do with emotions and totally whacko notions related to religion, political conviction, and such. If we decide that emotions are trivial formal systems, we humans escape the restrictions discussed here.

Re:Hmmm.... robotics?

[Anonymous Coward]

VIZZINI

But it's so simple. All I have to do is divine it from what I know of you. Are you the sort of man who would put the poison into his own goblet or his enemies? Now, a clever man would put the poison into his own goblet because he would know that only a great fool would reach for what he was given. I am not a great fool so I can clearly not choose the wine in front of you ... But you must have known I was not a great fool; you would have counted on it, so I can clearly not choose the wine in front of me.

THE MAN IN BLACK

You've made your decision then?

VIZZINI

[Happily] Not remotely! Because Iocaine comes from Australia. As everyone knows, Australia is entirely peopled with criminals. And criminals are used to having people not trust them, as you are not trusted by me. So, I can clearly not choose the wine in front of you.

THE MAN IN BLACK

Truly, you have a dizzying intellect.

VIZZINI

Wait 'til I get going!! ... Where was I?

THE MAN IN BLACK

Australia.

VIZZINI

Yes! Australia! And you must have suspected I would have known the powder's origin, so I can clearly not choose the wine in front of me.

THE MAN IN BLACK

You're just stalling now.

VIZZINI

You'd like to think that, wouldn't you! You've beaten my giant, which means you're exceptionally strong ... so you could have put the poison in your own goblet trusting on your strength to save you, so I can clearly not choose the wine in front of you. But, you've also bested my Spaniard, which means you must have studied ... and in studying you must have learned that man is mortal so you would have put the poison as far from yourself as possible, so I can clearly not choose the wine in front of me!

THE MAN IN BLACK

You're trying to trick me into giving away something. It won't work.

VIZZINI

It has worked! You've given everything away! I know where the poison is!

Re:Hmmm.... robotics?

[E 99]

But we also know that intelligence is not material in nature,

No, you don't "know" any such thing. Barring the possibility of projects so black we've never even heard of them, no one seems to know what intelligence is. Least of all you or I. Claiming you know its nature at this point in the development of science is absurd.

The only way that's absurd, is if you take it to be absurd that someone else has experienced something of consciousness that you have not yet experienced. The state of humanity's understanding of technological matters is straightforward to assess, because those on its leading edge publish what they know in journals and publicize it. That is not the case we the state of our understanding of consciousness. Science doesn't know how to study it, as it's not in the realm of science. There's no scientific evidence that it exists, except our subjective experience of it. If you're satisfied with the explanation of human behavior as the response to stimuli from electrochemical neural networks, then you have an explanation which has no need of a concept of consciousness. Material events caused by material events -- consciousness is superfluous to the theory.

Likewise, your idea of intelligence, as far as I can tell, is what you recognize in behaviors that is evocative of your subjective experience of that aspect of consciousness. If intelligence and consciousness only have meaning in relationship with our subjective experience, (as is argued to be the case with God) what place can it have in science, and on what basis, other than a faith-based basis that seeks to discount non-material origins of things -- can we assign it a material origin?

Re:Hmmm.... robotics?

[Eivind]

The only way that's absurd, is if you take it to be absurd that someone else has experienced something of consciousness that you have not yet experienced.

That's not knowledge. That's religion.

Re:Hmmm.... robotics?

[Eivind]

We may not hide results, but there sure is publication bias -- an experiment with a surprising and/or interesting and/or positive result is much more likely to get published than the oposite. There's a lot of science of the type: We tried X, nothing happened, going on. Unfortunately it seldom gets published.

Re:Hmmm.... robotics?

[Eivind]

Playing devil's lawyer here. How exactly, do you go at modeling when a specific atom of an unstable element will, or will not, decay ?

We believe we have a reasonable understanding of why and how it does so, but we also believe that there *is* no explanation for exactly when a particular atom decides to decay, or not.

Re:Hmmm.... robotics?

[KlaymenDK]

But to match the resolution of a human eye, you would need a 100 Megapixel floating point framebuffer Are you sure about this? You(r post) seem(s) bright and well-informed, but I believe human vision uses a powerful combination of high-def focus vision, lo-def peripheral vision, and a "memory buffer" to create the illusion of overall high-def vision.
There was some research on that sort of buffered vision (I can't find the link now), and I'm pretty sure the actual "megapixel" value is closer to 17 than to 100.
So basically you could start with the 17mp image to create your initial "sky", "shore", and "window" segments, then focus on each separately to refine their content. So instead of "parsing" the entire scene as one, you split them up and refine -- similar to where real-time raytracing is going (except the other way around, building an object model from a scene).

Re:Hmmm.... robotics?

[12357bd]

Visual search engines can be build very differently as you propose. The series of tasks you said (segmentation, texture propcessing, etc) are just 'logical' or 'simple' ways to treat image recognition. there are other models. The problem with image information is that:

1) is poorly defined. What is a 'shape'? for what is worth?, and

2) has a very high level of 'semantic' content, Everyboy knows what a 'face' or a 'car' is, but we cannot reduce this information to a simple set of shape/textures, there's no direct link between human 'objects' and graphical images, that's why image search by content fails so easily.

Without addressing the first point (mathematical image information definition) there's little hope for image search engines.

Re:Hmmm.... robotics?

[leonem]

Good on you for raising this. My own perception of the problem you describe with assuming humans are not Turing machines was much more vague; it mostly relied on my own feeling that, despite having read Goedel's theorems and learnt some of the methods of formal logic, I didn't really know it to be the case. As with your initial disclaimer, I think one tends to assume more intelligent people than you really have a grasp of it beyond yours.

Oddly enough, this returns to another poster's assertion that they had had a higher experience of consciousness than someone else; whether from a religious person or a mathematician, such statements must be treated carefully. This further returns to an earlier discussion about the place of philosophy in these discussions. I try not to believe anything in an absolute sense beyond the assertion, provable only to myself, that something exists. I can't prove it's 'me' ('I' might just think I'm me), but I am convinced that for any perception of anything to occur, something must exist.

Science, I think, needs only one assumption beyond this to be useful, which is that the world you observe has a consistent set of rules that will not change arbitrarily. Note, it doesn't assume the whole thing isn't an hallucination of some sort - it makes no comment on this - it simply tries to determine the rules of what is observed, with the assumption that they will not arbitrarily alter.

I think a lot of discussions on Slashdot conflate these quite different areas of thought, and the types of proof each one entails, and the domains in which such proof claims to operate.

Re:Hmmm.... robotics?

[fbjon]

What are you afraid of? The devaluing of humans, perhaps. No-one wants to be superfluous, so we want to consider ourselves magically above "the rest", or untouchable.

Re:Hmmm.... robotics?

[fbjon]

A tank can easily be recognized by it's silhouette, it's a simple 2D shape. A trash can however, can take many shapes, but they all have a hole in the top. Thus we have a system for recognizing 3D objects based on the 2D image we perceive (3D, or stereoscopic recognizing of a shape is very often inadequate). This is based on simple rules like geometry (vertical lines, circle or oval on top), which way shadows fall (curved and slanted shadow inside circle), etc. That the rules are simple is obvious, as the human eye is entirely trivial to fool with optical illusion.

On top of this comes pattern recognizing, like a grid-like structure (a plastic thrash can), curved reflection of surroundings (a metal trash can), non-shadowed texture within top circle (aha, no hole so not a trash can after all!), and so on, and so on.

Re:Hmmm.... robotics?

[fbjon]

If you buy the theory that humans originated in Africa, and developed our minds in order to survive on the savannah, then seeing circles, triangles, and squares is a useless ability. OTOH, lines are absolutely vital.

All of those species of trees you mentioned are practically identical: two roughly vertical lines (possibly curving) close together that have a contrast and texture change, compared with the surrounding. The underbrush is similar: 1000 species, all a green, unrecognizable mass except for brightly coloured berries and flowers, which stand out purely because of contrast difference and colour. The sky is what you find above the horizon, no need to analyze clouds to find that. "Above" is defined by a combination of gravity (feet point down) balance, and possibly a straight horizon, if available.

Panthers? They're camouflaged, so you have to do guesswork to find out where they're hiding (tall grass, which is a fairly even texture stretching out below the horizon).

Edible trees? Find one characteristic (shape or otherwise) of the trees you like, then look for those.

A large part of our basic vision needs is based on simple geometrical shapes like lines, circles, ovals, triangles, etc., the direction they're pointing, and shadows that objects cast (themselves geometrical shapes, frequently triangular-ish). A triangle doesn't need to be perfect in order for you to recognise it as a triangle. It just needs three pointy bits, and there are plenty of objects like that in nature. A "mathematical" triangle with straight edges is just a special case of a "natural" triangle.

Re:Hmmm.... robotics?

[lawpoop]

No - it doesn't have to mean that at all. Unfortunately, for our present situation, it does.

There may be other ways - or many other ways - to solve these problems that do not use the same mechanisms that human physiology does. That's probably true, but right now, we have only *one* way to emulate intelligence -- the Turing machine. So unless the Turing machine is basically a mind ( and I argue that Goedel showed that it is not ), then we have to build a new kind of device. It doesn't matter whether this other device is the same kind of thing as a human mind or not -- right now we don't have anything other than a Turing machine, so we have to build a new kind of device if we want some of the behavior that a Turing machine is incapable of.

Re:Hmmm.... robotics?

[lawpoop]

I think if what you were saying is true, we would have good computer vision, because we have a lot of powerful math for dealing with triangles, circles, and squares, and really powerful computers that can do that math. However, basic recognition tasks have still not been done by computers.

No tree looks like a triangle. From any angle, they have a cloudy, amorphous, fractal structure. There aren't any 'points' in the shadow, nor even straight lines. You can't understand clouds using classical geometry. You can't know anything about a river from it's shape using geometry -- you need fluid dynamics modeling.

A triangle doesn't need to be perfect in order for you to recognise it as a triangle. It just needs three pointy bits, and there are plenty of objects like that in nature. A "mathematical" triangle with straight edges is just a special case of a "natural" triangle. I'd like to see the computer program that can look at a tree's shadow and call it a 'triangle'. Hell, I'd like to see the computer program that can pick a shadow out of the grass!

Re:Hmmm.... robotics?

[lawpoop]

You are right, but my point is that far and away, what our visual system evolved to perceive were complex, non-geometric systems, such as cloud formations, or where the 'path' is as it wanders from the forest to the open field.

Re:Hmmm.... robotics?

[Jim Hall]

[...] robots that would be able to clean your floor, carry your groceries, navigate in a burning building, walk your dog, tend your lawn.

This is an important distinction for a house robot to make. I, for one, would not want my robot trying to take my azalea for a walk, or trying to prune my terrier. :-)

Re:Hmmm.... robotics?

[DragonWriter]

It may not be deep, but it does have serious implications for AI and computers. It means that if you want to have a conversation with a computer, or have true face recognition, or solve any hard-AI problem, we have to invent a new type of machine first.

No, it doesn't.

We don't have any computing device that's *not* completely a Turing machine.

Any existing computer that has both inputs and outputs to the outside world, where the outputs can affect the inputs through intermediary physical processes that cannot be completely modeled by a Turing machine, is, if you take the whole system of interactions into account, a computing device that is not a Turing machine, even though it contains one.

We don't have any 'Turing-Plus' devices;

Actually, we have quite a lot of them. You probably have several in your home.

Re:Hmmm.... robotics?

[lawpoop]

Gödel's theorems have nothing to do with representing the human mind in any form. Godel disagrees with you:

One of the earliest attempts to use incompleteness to reason about human intelligence was by Gödel himself in his 1951 Gibbs lecture entitled "Some basic theorems on the foundations of mathematics and their philosophical implications".[1] In this lecture, Gödel uses the incompleteness theorem to arrive at the following disjunction: (a) the human mind is not a consistent finite machine, or (b) there exist Diophantine equations for which it cannot decide whether solutions exist. Gödel finds (b) implausible, and thus seems to have believed the human mind was not equivalent to a finite machine, i.e., its power exceeded that of any finite machine. He recognized that this was only a conjecture, since one could never disprove (b). Yet he considered the disjunctive conclusion to be a "certain fact". [Emphasis mine]

I don't really understand the math beyond a metaphorical level. I don't really understand what completeness or inconsistency is. I don't think it matters that the human mind may or may not be complete or consistent. The point, as I understand it, is that a human mind is able to see the proof that a system cannot both be consistent and complete, whereas a Turing machine would never be able to demonstrate that.

However, referencing Gödel's incompleteness theorems just because they sound appropriate at first glance does not give any argument scientific credibility. Similarly, referencing a book about Goedel's theorem just because is might be relevant has no scientific. I have no reason to believe Torkel Franzen over J. R. Lucas or Goedel himself. It all depends on the details of their arguments. In this case, Lucas is more convincing to me. A human being can 'understand' that a Turing machine can never understand certain things. It doesn't matter that a human can't necessarily understand those things either; all that matters is that a human can understand that a Turing machine can't understand something, which is a feat that the Turing machine can't do.

Re:Hmmm.... robotics?

[lawpoop]

Actually, we have quite a lot of them. You probably have several in your home. Can you give me a few examples? Are they ones that depend on input from human beings?

Re:Hmmm.... robotics?

[DragonWriter]

Can you give me a few examples? Are they ones that depend on input from human beings?

Most computing devices used by humans "depend on input from human beings", however, any that have any loops between an output and an input that include a process that isn't equivalent to a Turing machine are Turing-plus without considering the human interface, even if we don't particularly have any idea to apply that to general reasoning. Heck, a modern computer where the operating system or BIOS (any software that the CPU runs, at any rate) interfaces with a heat sensor to monitor the system's temperature and control fan speeds is Turing-plus, since the output it sends to control the fan affects the input it receives on the heat sensor through a process that cannot be fully modelled by a Turing machine.

Of course, we don't apply that process for generalized reasoning, and that's true of most such output-input loops, they are usually parts of fairly-specialized control subsystems. But that's a matter of understanding ways to apply the non-Turing features that are inherent in such systems, not a matter of our present computing devices being strictly Turing machines.

Re:Hmmm.... robotics?

[mikael]

Here's a couple of articles I read:

Facts about the brain
Rods and Cones

There are around 125 millions rods and 6 million cones in each eye, with the percentages of each color/wavelength (red = 64%, green=32%, blue=2%)

No Sense

The human eye has 100 million neurons per per eye of five types, but there are only around 1 million neurons per optic nerve (arranged in bundles of 1000).

Re:Hmmm.... robotics?

[DragonWriter]

No, we religious types know that intelligence can be created. But we also know that intelligence is not material in nature, and therefore that it can't be built out of material substance.

Not all of "we religious types" believe that, and arguably none could know it, even if it were true (but that quickly descends into a thorny epistemological debate.)

I don't think we coined the term though. The term was probably coined by graduate students researching with neural networks, who wanted it to sound to the girls like they were on the verge of building HAL 9000.

"Strong AI" and "Weak AI", as I recall from my Intro to Philosophy class years ago, were coined by John Searle, a philosoper. IIRC, the initial distinction was essentially metaphysical (i.e., Searle argued, in effect, that even if it could pass a Turing test, an AI could not be 'really' thinking and would still not be a 'Strong AI'). Though "Strong AI" vs. "Weak AI" is now used often to distinguish, in a very fuzzy way, the level of generality of an AI rather than the "really thinking" distinction, and frequently the two are conflated as if generality were equivalent to "really thinking".

Re:Hmmm.... robotics?

[aragszxki]

I should have been more precise in my statements, I apologize for being unclear. When I said "Gödel's theorems have nothing to do with representing the human mind in any form", I should have emphasized the word representing. I did not mean that the incompleteness theorem is philosophically unrelated to the topic at hand. I merely said (and I reiterate):

They cannot be applied to the human mind for the purposes of answering the question of strong AI. I do not want to argue if the mind is a Turing machine or not. I just want to state that there is no proof in Gödel's, Lucas' or Franzen's works for this issue. Assuming the truth of the statement "the human mind can understand something a Turing machine can't" is a fallacy.

I am also aware of the 1951 lecture. However, that lecture or anything else produced by Gödel does not prove that a human can understand something that the Turing machine cannot. It does not disprove of the fact either. You are welcome to believe anything you want, but that does not change the fact that the proof simply does not exist.

Re:Hmmm.... robotics?

[lawpoop]

I guess at this point it's a matter of definitions, but I would still say we don't have the device yet. You were right; I was wrong; we do have 'Turing-Plus' devices. However, we don't know *yet* what specific Turing-plus device can give of the behavior of the human brain. We can hook up cameras and thermometers to Turing machines, but somehow we haven't yet come up with AI. And I suspect that a Turing-plus device may not even be what gives us AI. I think we might get a device that can solve strong AI problems that contains *no* Turing machine. But, i also do believe that the human machine contains at least one Turing machine, so a complete re-creating of the mind would contain a Turing machine.

Re:Hmmm.... robotics?

[DragonWriter]

I guess at this point it's a matter of definitions, but I would still say we don't have the device yet. You were right; I was wrong; we do have 'Turing-Plus' devices. However, we don't know *yet* what specific Turing-plus device can give of the behavior of the human brain.

Well, yes, but that's a very different than only having available classes of devices that provably are inadequate to the task.

We can hook up cameras and thermometers to Turing machines, but somehow we haven't yet come up with AI.

Well, since we don't have a rigorous definition of what generalized AI needs to do (we have mostly negative tests, but even using our own Turing-plus minds we can't formulate a concrete description of the processing involved, and its not even clear that we could, short of actually developing AI, actually come up with such a description) we're essentially left with trial and error and attempts to mimic some of the smaller component processes without fundamentally understanding them except as black boxes. This is a hard problem where progress is understandably slow. But developments of machines that can do things that are components of intelligence that AI naysayers have, in the past, said machines could not ever do aren't that uncommon. General AI is, I suspect, always going to seem far off until and unless other limited AI technologies converge to form it; I don't think there will be a point where General AI will be widely seen as "imminent" before it occurs, because we won't understand the connections until they are made.

I'm not saying its likely at any time soon, though.

And I suspect that a Turing-plus device may not even be what gives us AI. I think we might get a device that can solve strong AI problems that contains *no* Turing machine. But, i also do believe that the human machine contains at least one Turing machine, so a complete re-creating of the mind would contain a Turing machine.

You seem to be using an odd definition of "Strong AI problem" here. If you mean the kind of problems that are conjectured to be "AI-complete" (requiring generalized human-level intelligence even if they appear to be narrower problems), then if the human machine includes a subsystem that can be modelled by a Turing machine, any machine that could solve such problems must necessarily do so as well.

But solving particular problems is what characterizes "Weak AI" in the original distinction vs. "Strong AI", which involves having characteristics like "consciousness" independent of the ability to solve particular problems.

Re:Hmmm.... robotics?

[lawpoop]

Oh, maybe now I understand.What you are saying is that Goedel's theorem has nothing to do with a device that can perform the tasks of the human mind, but is a fundamentally different type of device?

However, that lecture or anything else produced by Gödel does not prove that a human can understand something that the Turing machine cannot. It does not disprove of the fact either. You are welcome to believe anything you want, but that does not change the fact that the proof simply does not exist. OK, so if the proof does not exist, what do you think of this argument? (Hilary Putnam, _Minds and Machines_) It seems pretty straightforward to me. What am I misunderstanding?

Let T be a Turing machine which "represents" me in the sense that T can prove just the mathematical statements I prove. Then using Gödel's technique I can discover a proposition that T cannot prove, and moreover I can prove this proposition. This refutes the assumption that T "represents" me, hence I am not a Turing machine. Can a Turing machine demonstrate Godel's incompleteness theorem? Can a human always stay one step ahead of a Turing machine, insofar as what propositions can be proven?

Re:Hmmm.... robotics?

[aragszxki]

I'm sorry, I did not exactly understand your first sentence. I'm not sure if this is what you were asking, but I'll try to explain. What I'm saying is that Gödel's theorem is a purely mathematical theorem that is only relevant to formal systems strong enough to support basic arithmetic. Anything it proves cannot be used to support facts outside the domain of arithmetic. One can be inspired by what they imply, but that does not equal proving those statements.

The argument you have mentioned makes clever use of circular logic to misdirect the reader. There are two possible cases:

(1) If T can prove the mathematical statements I can prove, it can use Gödel's technique just as well as me. After all, they are mathematical statements as well.
(2) If there exists a statement that T cannot prove but I can, then T does not represent me totally. However, I can always build another Turing machine, T-Prime, that can prove everything T could prove PLUS the statement that T failed to prove. This machine can now be said to represent me totally.

I'm sure you see where this is going. Use the incompleteness theorem to go beyond the limits of this machine, build T-Prime-Prime to solve that, use the theorem once more, build T-Prime-Prime-Prime to solve that and so on. There is no end to this recursion. The resolution "...hence I am not a Turing machine" cannot be reached. The resolution "I am a Turing machine" is not reached either. Once again, the proof is not there.

Can a human always stay one step ahead of a Turing machine, insofar as what propositions can be proven? As I've explained above, it is not the human that is always one step of the Turing machine. They are not ahead or behind each other in any way. The recursion is simply infinite.

Re:Hmmm.... robotics?

[lawpoop]

The argument you have mentioned makes clever use of circular logic to misdirect the reader. No, it's not circular reasoning, and there is not misdirection. Goedel's theorem is recursive. No matter how powerful the system, there are always truths outside of that system. If you build a more powerful system to show the truths that the weaker system couldn't, there are more, *other* truths that the newer, more powerful system cannot show. The human is aware that there are truths that a Turing machine cannot show, but a Turing machine cannot never show that there are truths that it cannot express. It's not a trick of 'circular reasoning' as you claim.

(1) If T can prove the mathematical statements I can prove, it can use Gödel's technique just as well as me. After all, they are mathematical statements as well. This is they key. I think Goedel's theorem cannot be shown by a Turing machine. In other words, while a turing machine can do a lot of math, there is at least one kind of math that a Turing machine cannot do, but a human mathematician can.

(2) If there exists a statement that T cannot prove but I can, then T does not represent me totally. However, I can always build another Turing machine, T-Prime, that can prove everything T could prove PLUS the statement that T failed to prove. Good so far...

This machine can now be said to represent me totally. Wrong! The property of the human mind that we are interested is not the statements that you can prove that a particular Turing machine cannot, but the ability to apply Goedel's theorem to show that there are statements that a Turing machine cannot show. the Turing machine doesn't represent you because it cannot apply Goedel's theorem to itself, not that it can do all the proofs that a weaker machine couldn't. The ability to apply Goedel's theorem is the ability of the human mind that shows it is different from a Turing machine.
You can re-apply Goedel's theorem to create a statement that you can show, but the newer T cannot. Therefore, newer T is does not represent you totally. No matter what you are always capable of applying Goedel's theorem to make a statement that a Turing machine cannot. The Turing machine is incapable of applying Goedel's theorem.

I'm sure you see where this is going. If it is true that I can see where this is going, then I am not a Turing machine, because a Turing machine cannot see where this is going. I am aware ( if I were a sufficiently powerful mathematician that understands Goedel's theorem) that there will always be truths that a Turing cannot express, no matter who powerful it is; that Turing machine will never be able to show that there are truths that it cannot express. Sure, another Turing machine can expressed truths that a weaker one can't, but no Turing machine can ever show that there are truths that it itself cannot express.

Use the incompleteness theorem to go beyond the limits of this machine, build T-Prime-Prime to solve that, use the theorem once more, build T-Prime-Prime-Prime to solve that and so on. There is no end to this recursion. So, is the Turing machine able to show that there are truths that it cannot show? If not, a human can do something that a Turing machine cannot. Therefore, the mind would not be a Turing machine.

The ability that the human mind has that the Turing machine does not is the ability to perceive this recursion, to understand Goedel's theorem.

As I've explained above, it is not the human that is always one step of the Turing machine. They are not ahead or behind each other in any way. The recursion is simply infinite. I think this is the catch. The entire recursion can be conceived or expressed in the human mind, while a single Turing machine cannot show the infinite recursion. You actually have to build an infinite series of Turing machines to completely express the recursion. It seems that you are claiming that the human mind is an infinite series of Turing machines. This would mean that the mind is not a Turing machine.

Re:Hmmm.... robotics?

[aragszxki]

In the light of this post, I feel like we have entered a Gödelian recursion, repeating the same things at different levels. So I will stop here and concede. I am actually a researcher working on strong AI so I already spend my working hours poring over these issues. I'd like to keep some of my free time actually free of work. Thank you for a stimulating discussion.

Re:Hmmm.... robotics?

[lawpoop]

Thanks for giving me the chance to discuss this with someone who knows what they are talking about. I will check out the book you recommend! :)

Re:Hmmm.... robotics?

[fyngyrz]

The only way that's absurd, is if you take it to be absurd that someone else has experienced something of consciousness that you have not yet experienced.

No. It is absurd from beginning to end. Your mind can produce any experience you dream up. It can provide the feeling of time stretching, imaginary sexual encounters, terrifying monsters, cobble up situations that never existed at all, tweak you with false recognition of events, mis-identify people and objects, drown you in beliefs from Islam to Voodoo and Astrology and everything in between, it can convince you that you were molested when you weren't, convince you that you were not loved when you were, that Santa is real and government is benign, and so on for pages. It can do this awake, asleep, with or without the assistance of pharmaceuticals. The fact is, the mind is very rarely able to perform as a reliable measuring instrument without external forces to keep it in check.

The strength of science is that it provides a reliable, repeatable, consensual, standardized means for looking at reality outside the boundaries of our skulls. If you say "Do X, Do Y, with Q and R, using protocol M, then Z happens and so I think this may imply..." we can follow you with Q...Z and then take a dispassionate, consensual look at your idea and see if everyone's experience agrees, and other confirming tests can be found. When you say "Man, I had this experience last night, the Lord came and told me consciousness is immaterial", that's not evidence of anything but random junk rattling around in your head. Likewise if you pull a reference out of a book written by those people driven by belief, instead of knowledge.

Likewise, your idea of intelligence, as far as I can tell, is what you recognize in behaviors that is evocative of your subjective experience of that aspect of consciousness.

No. I told you already; I don't know what intelligence is. Please don't imply that I've defined it, or attempted to do so. I have not. I am perfectly comfortable with questions that are not answerable. I find the questions interesting, that's all. Attitude-wise, I don't make things up to fill a silence; I wait for an actual sound.

Re:Hmmm.... robotics?

[fyngyrz]

Unfortunately it seldom gets published.

It is more likely to be if it upsets an applecart, though. You can't publish everything; every high school and college reproduces certain basic experiments and we'd be drowned if every confirming result was published. For instance, when Pons and Fleishman made their announcement, all kinds of results were reported, and shared less formally amongst labs and researchers.

I'm considerably more concerned with political manipulation of results than I am with the publication bias driven by level of surprise or interest. Political from both ends; the loudest shouters from office and the rank and file always seem to be the least qualified [grump.]

Re:Hmmm.... robotics?

[fyngyrz]

...we want to consider ourselves magically above "the rest", or untouchable.

Tonto and the Lone Ranger were hiding in a stand of trees; Geronimo was circling with a thousand well armed and very angry warriors. The Lone Ranger turns to Tonto and says "Well, I guess we've had it, Tonto." Tonto looks him right in the eye, and says "What we, white man?"

Re:Hmmm.... robotics?

[fyngyrz]

However, statistically, we know how they behave in groups. So a great deal of successful and useful modeling can model the behavior of the group, then randomly choose victim atoms to force conformance with the known group behavior.

Back to the individual atom: Remember I said that we can model for systems we understand; you say that in this case, there is no reason for a particular atom to decay. This means that the decay of an actual atom in a system is beyond understanding. There is no contradiction here.

Re:Hmmm.... robotics?

[shiba_mac]

Firstly, my apologies for taking a long time to reply. I'm afraid we had a power outage at my house for a few days, and when it came back, this had slipped from my mind. Thank you for taking the time to respond.

You seem to imply that you don't see a big difference between religion and philosophy. I can see where you are coming from, but I feel you are mistaken. Religion, or Theology at least, is some sort of philosophy, but philosophy is not a religion. Lots of philosophers believe in god, or some sort of "first cause". Lots don't. Lots of scientists believe in god. Lots don't. It is possible to be a believer in God/Allah/Buddha/The FSM, and to be a good scientist or a good philosoper. If someone claims to be a member of a church based on the teachings of a particular philosopher then they either haven't read widely enough (or at all), or they had a bias when they started to study.

I would argue also about the very sharp line you draw between philosophy and science. "Way back when", science was done by philosophers, as you say yourself. I'd argue that it still is, and we've just had longer to get better at it. The inventor of calculus was a philosopher and a scientist. Science is a philosophy. A philosophy where experiments (provably correctly) are used to investigate hypotheses. But still a philosophy. To be a scientist is to stand on the shoulders of the giants (and the normal sized people) of the past, and not to acknowledge this is at best ingracious.

You refer to the scientific method as a series of "mundane, easily understood and followed steps". There is nothing mundane, or easily understood, about the formulation of hypotheses. Many times the explanation will be obvious (how disease spreads, once you've invented the microscope), and many times it will not (why is Mercury's orbit around the Sun so funny looking). The scientific method is an example of something born of pure reason, which has no real existance except in the minds of people, Which has no power unless people believe in it. And yes, it does work.

pr0n

[tronicum]

great! new way to find even more porn.

Re:pr0n

[Sciros]

Yeah, what kind of queries are you going to use? Gonna buy a tablet so you can make crude drawings and then browse to them on a page and click "search" to find stuff that matches your drawings?

Man, I don't know about you, but I gotta work on my drawing skills...

On that note, female models interested in expanding their nude modeling portfolios please email me.

Re:pr0n

[grub]

Yeah, what kind of queries are you going to use?

1: Type pink -goatse in search box.
2: Click [I'm Feeling Lucky]

??? profit etc etc

Re:pr0n

[kebes]

You may be joking... but I think rather than submitting sketches, the user would submit samples of things that already match what they want. For pornography, if a search engine were able to find images similar to those already tagged as "I like this," that would be a really sought-after search engine!

More broadly, if a search engine were able to find similar pictures, then you could narrow down to the result you wanted by submitting images that are close to what you want. For instance you may have found a thumbnail of the image you're looking for, but can't find a full-size version. Or you have a few pictures of airplanes, but want a whole bunch more.

Another way the technology could work is to present you with a series of candidate images, and then you click on the one that is "closest" to what you want. It then performs the search again, showing new candidates, and you click on the closest match again. If the search engine keeps showing you things closely matched to your last few selections, then this iterative process would quickly home-in on images of exactly what you want. In this kind of mode, the search engine could be using visual similarity as well as keywords and tags to figure out what kind of images you're trying to find.

Lastly, the idea of sketching an image might also work--at least for simplistic images. For instance I've often thought that the symbol-picker applet should, instead of listing thousands and thousands of symbols (which font should I look in?), it should have a box where the user can clumsily draw the symbol, and then display close matches (if you draw a circle it would show the degree sign, the letter 'o', the number '0', etc.). If it works, image-similarity technology like this could be a way to find the desired symbol. (It might work for clipart, too.)

Re:pr0n

[Anonymous Coward]

How likely is it that Rodney King searches yield results of a different kind of gang bang?

Re:pr0n

[Anonymous Coward]

Or you have a few pictures of airplanes, but want a whole bunch more. Airplane porn? Damn, just when you think you've seen everything...

Re:pr0n

[Sciros]

Yah I was just joking around.

Really though, suppose you don't have any images of two chicks riding a wookiee in a gladiator outfit. And say you know there's one out there. Well, I'll tell you, Alex Ross has a much better chance of finding that image with his mad drawing skillz. Of course, once he completes his "query," he's made himself the image he was looking for. So I guess it's kind of pointless. I forget where I was going with this anyway.

Re:pr0n

[amchugh]

Well I guess we know now what a wookie would do if he got his hands on a million dollars.

You have to have seen office space for this to be remotely funny

Re:pr0n

[egoproxy]

it should have a box where the user can clumsily draw the symbol, and then display close matches

That is already possible. I have not specifically tested it with the symbol font but it works with asian fonts.

For example the IME in WinXP allows you to scrawl a character and it narrows down the matching choices as you draw. Unlike the Hitachi version, however, the IME matches based on stroke order and direction.

Re:pr0n

[Wiseman1024]

Rule 34.

Re:pr0n

[Wiseman1024]

That'll be amazing, I can't wait to try it. Their slogan could be:

"What do you want to fap to today?"

Re:pr0n

[zeylisse]

great! new way to find even more porn. Yeah, but can it tell the difference between hardcore pr0n and a picture of, for example, Fidel Castro eating a banana?

Re:pr0n

[Anonymous Coward]

What difference? :S

creative headline

[BendingUnit]

Calling this a search engine is a stretch. The article even calls it "search technology".

Nuthin but a B-tree of eigenvalues

[gatkinso]

and life ain't nuthin but bitchez n money.

Hash table?

[Simon80]

Sounds like a an on-disk format involving hash tables. They'd probably win a patent on it, too.

Re:HDD

[Anonymous Coward]

Usually, HDD = Hard Disk Drive

Will Google copy or buy this technology?

[bepolite]

I would think this would be a big and useful upgrade for http://images.google.com/

Re:Will Google copy or buy this technology?

[Anonymous Coward]

Mm. I used to expect Google were looking into similar image search, tune search and the likes, years ago, to expand from dominating text search to multimedia search. Then they started veering off into office web apps and the likes.
Still, they should actually be sitting on a team working with this kind of thing, with the recent Youtube acquisition and that copyrighted content detection thingie they were implementing.

Robots

[Odinson]

Yes, but how quickly can this be integrated into robots. Robots programmed to destroy all buttons.

"Kill multi button gadgets! Steve Jobs robot army angry!"

Re:Robots

[tehdaemon]

That would be a scary army indeed, you couldn't have an off button!, (would a switch do instead?..)

T

1 second...

[Anonymous Coward]

On what hardware? A 16MHz 386?

similar to Video Google?

[gstone]

From the rather less than opaque description in the linked article, it seems that this works is a hierarchical extension to a system known as Video Google. This system detects two-dimensional features in every image of a video sequence. Then uses hierarchical clustering to group together "like" features together. The centres of these clusters are used as "visual words". Scenes from the original video can then be characterised by which of these visual words they contain.

Using these words, search engine style indices and techniques can be used to make searching -- by supplying an example image area which can have its words computed -- quite fast.

The key bottle neck here is the clustering stage: reducing the original input of typically hundreds of features per frame -- multiplied by 25 frames per second by minutes, or hours, of video -- to a much smaller set of clusters. It looks like the work in the linked article is using a modified clustering algorithm which does not require all of the data to be in memory at once.

The TRECVID project is a challenge style exercise where groups compete to provide the best search results for a given set of queries where the search material is hours of video.

Document images

[zymurgyboy]

Pictures are fun, but I wonder if it would be accurate enough to locate similar images of documents (and to what degree). It would be really cool (for me anyway) if it could look at, say, a million pdfs or tiffs that don't have embedded text and come back with everything similar/identical.

I frequently have to create large collections of images from all sorts of file types -- some text-based, some graphics -- that get housed in a collection of images for easy, standardized review. If there were something that could avoid the step of extracting text from them, or later OCRing them and still end up with a searchable image collection, well, that would be exceedingly cool. It would cut the initial time outlay I have to devote to virtually any given project I have to deal with by 25 to 50%.

Re:Document images

[Anonymous Coward]

I frequently have to create large collections of images from all sorts of file types

Yeah, I collect porn too...

Re:Document images

[FleaPlus]

I frequently have to create large collections of images from all sorts of file types -- some text-based, some graphics -- that get housed in a collection of images for easy, standardized review. If there were something that could avoid the step of extracting text from them, or later OCRing them and still end up with a searchable image collection, well, that would be exceedingly cool. It would cut the initial time outlay I have to devote to virtually any given project I have to deal with by 25 to 50%.

Have you considered trying some SIFT-based algorithms? It usually performs pretty well on detecting matches between 2D forms, if I'm understanding your problem correctly. There's some open-source libraries that implement SIFT, as well as a pretty nice demo from Evolution Robotics that applies SIFT to data from USB cameras, so you can have it do things like recognize dollar bills, book pages, etc.

Re:Document images

[zymurgyboy]

No, I hadn't. Didn't know they existed until just now. Thanks! The demo you linked to advertises a device with an SDK, which may be worth looking into.

Something like this could certainly be another trick for the proverbial bag. The one area where something like this would fail is when it comes to concept-based searching, where you'd pretty much have to have the text to feed an algorithm. For instance, "in the dog house," and "in big trouble" have similar meanings to a (US) human, but a garden variety keyword search wouldn't hit for both of them unless you searched for trouble, dog and house. Some of the linguistic and concept based search technologies would catch turns of phrase without you having to specifically search for the different variances of the same concept. That's pretty neat until you throw in a bunch of photographs in the same collection, at which point text searches are little if any help at all.

Double-byte characters are another dimension of the same problem, squished together if you will. Most of the software commonly in use (which means software I have to use that runs on Windows) for my purpose has little or no support for it. You may be able to extract the text and load it in a database, but you won't be able to search on it. Just ran into that today. That's a limitation of the present state of the software I'm using that will probably be dealt with in the near term. But honestly, I don't read any double-byte languages, so I'd probably be just as well off with an image-based comparison. Especially if it were faster than a text search that yielded the same results. Something like this would be killer for deduplication.

Re:Document images

[12357bd]

If what you need is just image comparison, without textual information processing there are solutions (ie: http://www.immem.com/en/). What do you mean for 'avoid the step of extracting text from them, or later OCRing'? Are you doing text or image comparisons?

Re:Document images

[zymurgyboy]

I'm making images of different types of files. If there is text in the original file (word processing files, spreadsheets, slide presentations, etc.), that gets extracted during the imaging process and loaded into a database for searching. Some things don't have embedded text but have a textual component of some sort (think jpeg photograph of a stop sign). Or I'll get documents previously imaged by someone else and don't get OCR or extracted text. In those two cases, the image is OCRed and the OCR text is loaded into a database to make the image collection text searchable.

What would be nice would be to have both an image comparison and a text search. Text search is good for a lot of things, but falls down when it just isn't there and/or can't be generated. In that case, an image comparison would possibly fill the gap. And faster is always better.

Re:Document images

[12357bd]

Image comparison can be made real fast if it's purely a graphical comparison (ie: detect duplicate/resized images). The problem is when you need to search for 'content' (ie: search 'houses' or 'cats') then the cognitive aspects of the problem results in severe limitations of the different search motors/technologies, each one has his own set of strengths/weaknes.

This is going to spell the death...

[saveourskyline]

...of 48yr old, 365 lb. pound guys who steal pictures of girls from MySpace and pretend to be 14/f/kali.

Re:This is going to spell the death...

[zmollusc]

Oh noes!! (desperately tries to cram cat back in bag)

Not as useful as it sounds...

[Kirin Fenrir]

The technology can't determine what aspects of the image you're looking for.

For example: I want to find more cat images. I feed it a picture of a white cat. I am more likely to be returned results of white dogs than, say, tabby or black cats.

Unless I'm misunderstanding something?

Re:Not as useful as it sounds...

[Jrabbit05]

But it could be used to create algorithms to find quality pictures, good photographs without viewing all of them.

Re:Not as useful as it sounds...

[fyngyrz]

...add one tool, a freehand outliner to select the most important aspect of the sample image.

This allows you to select the white kitten from the rest. If this technology can't tell a kitten from a puppy, it is pretty useless anyway.

Re:Not as useful as it sounds...

[FleaPlus]

For example: I want to find more cat images. I feed it a picture of a white cat. I am more likely to be returned results of white dogs than, say, tabby or black cats.

It seems it would be straightforward to implement something analogous to Google Sets, where you could supply a few photographs of what you're interested in (say, several cat pictures of various colors, or several white-colored pets). It could then learn which of the features were relevant, and add weigh to those in its search.

HDD-Take a load off.

[Anonymous Coward]

"What is this HDD they speak of repeatedly?"

Heavy Dump Diapers.

Hmm... FBI contract with Hitachi, maybe?

[jamestheprogrammer]

I could see the FBI paying some millions of dollars for a dedicated system like this... I mean, since they have that known terrorist photo database or whatever, they might want to improve performance... Of course, I would hope that the FBI would properly configure the servers if they were to buy this. They accidentally forget to change the server from images.google.com (or something similar) to terrorists.fbi.gov, and all of a sudden, your granny is a known terrorist. Oh no!

Re:Hmm... FBI contract with Hitachi, maybe?

[dj_tla]

The FBI (law enforcement in general) already has pretty sophisticated systems for face recognition and matching fingerprints, which is really what they're concerned about. A general puropse image matching algorithm will never fare as well as one tailored to a specific problem.

The answer is obvious

[CanadaIsCold]

If we deployed a large solar array in space and then used it to generate energon cubes then we could just use the shuttle to collect the energon cubes. Later after we gathered enough we could build large transforming robots to collect the cubes. These robots would be powered off the cubes as well.

Since when can an ancient indian tribe ...

[SengirV]

... long since forgotten, be responsible for such innovative technology?

Re:Since when can an ancient indian tribe ...

[Anonymous Coward]

Mod parent +1 Simpsons reference

Re:Hmmm.... robots scare me.

[Anonymous Coward]

Consider the scope of comparison - that's a lot of data to crunch, regardless of tableture.

If you have a PERFECT match, I'd think it would be straightforward, performance not withstanding.
What happens if the image got resized, cropped, pixelated, format/attributes, etc? Fast, but robust?

If you look for a license plate, (as they do now) the range is very specific.
If you want to find the guy with the fake beard and a limp, it gets tricky fast.

Time to update my disguises so the robots don't know who they're crushing with their giant metal claws.
B*n.Lee, I hope you witness the death of privacy and are appropriately wistful. Cheers dude.

Hmm, sounds familiar...

[Space cowboy]

... and the theory behind what I was doing is up at my blog. Or at least most of it is (all of it modulo time constraints. It'll all get there eventually).

Back in the day (almost 2 decades ago), I was using video rather than still images (which allowed me temporal information as well as spatial information) but I recently wrote a simple application to just use the spatial information to find me images "most-like" a source one. The original goal was to train the system and then try to leverage a semantic processor from the trained system. It worked reasonably well (sometimes astoundingly well) on the database I had (some 300,000 images downloaded from keyed-searches on google images).

As Hitachi said, the key is to develop a matching system within a higher numerical dimension. One of the missing pages on the blog (I'll get there!) is how to evaluate the usefulness of any given feature (=dimension) of a region of an image. With this, one can approximate a numerical value for the information being relayed to the recognition-system using that feature, and therefore establish its worth as a feature.

When you know what you're looking for (your feature set) *and* the value of each of those features to your recognition system targets {man,boat,grass,house,...} you can create reasonably useful discriminators and rule-systems based on those discriminators. Note that the discriminators and the rule-system can be given to the system as a-priori information, but most of them are created and destroyed automatically *by* the recognition system as it evolves. It sounds complex, but really it's a bunch of simple ideas applied one after the other.

Simon.

Desktop search with image pattern recognition AI

[backslashdot]

How long before you'll be able to search through pictures or video and the computer does image pattern recognition. So you can type the word "beach" or "jogging" and it will show you all the pictures showing scenery of a beach (or jogging .. or err jogging on the beach). Since camera makers dropped the ball and don't have easy intuitive image tagging capability built into the camera. Ideally a camera would have by now had voice recognition or recording so that you can tag a photo like "me in front of eiffel tower in paris" prior to taking it. Or at the very least a touchscreen system with common options. So now it's up to the AI of the computer to figure out the content of images using mad ocr level image analysis technology. Hmm maybe I should patent something like that.

Could this technique be applied to sound files?

[mikeasu]

Example - my brother burnt me a CD a while back with an irish instrumental I just love. No idea who it is, haven't heard from him yet about it. I was thinking it'd be neat to be able to search for say, a match to maybe 10 seconds of the chorus.

Re:Could this technique be applied to sound files?

[Anonymous Coward]

Doesn't every cell phone company in North America offer a service like you describe?
Maybe a Google search for music identifier?

Re:Could this technique be applied to sound files?

[pyropunk51]

You could try http://dump3.sourceforge.net/. It needs some work but the basics are there!

Re:Could this technique be applied to sound files?

[Anonymous Coward]

Yup it's been around for a while now... it's called SHAZAM.

In a nutshell, you dial 2580 when you're out, and then when the service gets a match it hangs up and texts you back the name & artist of the song. Quite neat.

Re: Godel

[Anonymous Coward]

I have been sitting on a proof that there exists a set of laws of physics for which the Turing Machine halting problem can be solved. It is also possible in that set of laws for the halting problem to be solved for itself.

I wonder what that does to Godel's theorem.

Re: robots, kitchens, and shipyards

[Anonymous Coward]

But if the thing is in a kitchen, why would it need to be looking to recognize images that are found in a shipyard?

Duh, because like what if your kitchen window overlooks a shipyard? The robot might look out the window and be confused if it didn't have a database of things found in a shipyard!!

However, on a more practical note, suppose as before your kitchen overlooks a shipyard. Did you ever stop to think that maybe a runaway freighter might crash into your kitchen? I betcha didn't think about that, Mr. Smarty pants. If you don't program the robot to recognize a ship crashing into a building, then it won't know it needs to try to unplug the toaster and salvage your pop-tarts before they get wet! Sheeesh.

Kids these days. Next you'll be asking me why a robot would need to know what sharks look like...

Saw it done 10 years ago

[mattr]

Some time between 1992 and 1994 IIRC when I was working at the photo/press agency Pacific Press Service in Tokyo, I saw a demo of a system created IIRC by NEC which searched 90,000 photos in under one second, based on a color freehand drawing you would draw on the screen of the EWS unix workstation on which it ran. Basically if you drew a horizontal blue mass at the bottom of the screen you would get a lake, etc. In other words you could search by rough photographic composition. I am less impressed that after over 10 years Hitachi was able to do something along the same lines.

Re:Saw it done 10 years ago

[egoproxy]

I've tried this but instead matching a lake I get:
blue sky
blue bus
bluish sunset
snowy mountains
a flower (with blue background)
etc...

I wonder if the aforementioned NEC program was any more useful...

Re:Saw it done 10 years ago

[dr_blurb]

do a search on "content based image retrieval", and maybe "survey",
and find dozens, if not hundreds, similar systems

Re:Saw it done 10 years ago

[mattr]

Hi!

I am not aware of it being used for a real project. It wasn't perfect either. At the time I was researching digital tech for photo agencies in Japan and I was more interested in natural language English text searching by Picture Network International, we were told the same people made a database for the White House. But they wanted to take over the industry with online contracts and got booted. For real use a foxpro system on a 386 was more useful. But the reality is that professionals in a library with endless drawers of color slides (20 per plastic page IIRC) had trained their minds to be superior to this system. You would learn about the library and where things generally are, and then pull out armfulls of plastic sheets over to lightbox tables. Basically you could look at a page of 20 slides in a second or two and know if a composition you want is in it. So no, it seems to me that image composition algorithms are (especially if pushed) much faster than what you might expect, but their practicality in the real world is limited.

Absoultely Wrong, analog != digital

[TheCouchPotatoFamine]

You may tell me that a turing system could emulate the brain, and i could tell you that a lighter could melt a bridge. A turing machine is meant compute in distinct sequences; no matter how many cores your cpu is running some basic algorithms will always have an sequence of steps.

The answer to real intelligence has zero-zip to do with sequential calculation. Here's a (ahem) parallel:

an analog delay pedal for music is a few dollars worth of tape and circuits that anyone can build in their living room. it has no latency of calculation, actually, the exact algorithm was never expressed in mathematical form at all!

The reason why turing machines will never emulate neurological systems effectively (even if you make a silicon-laser based neural network, let's say) is that they do not calculate their result as the SIMULTANEOUS calculation that a physical property (the addition of light, the addition of electricity)

And we know that reality is discontinuous at um... Plank size.. but computers are discontinuous by whole numbers (it's all binary in computers, whole binary numbers) and to *calculate* the number that represents the summing of say, 15000 other nodes is going to exorbitantly increase and increase.. and be a rat race to nowhere.

Moral of the story,
        - DIGITAL aka sequential discrete algorithms will always be out of reach
                          the length of time it takes to solve a problem increases exponentially
                          faster with problem difficultyblooms with the ANALOG aka physical
                          continuous summation

        - ANALOG aka intantaeous continuous algorithms based on real-world phenomena
                          scales linearly with the diffculty of the problem. I can't really say that -
                          no one really knows how the brain scales in that regard, but I can certainly
                          say that a whole class of problems exist that will forever be out of the reach
                          "classical" sequential machines.

What some people are thinking is right though, digital computers can do things that brains can't, and they'd be right. We've spent along time with digital logic and found alot of it's basic theorems and limits.

To close: It's the other side, the continuous field of analog electronics - not turing! - that has the promise of creating the type of intelligence we equate with "intelligence" and "free will". /"Free-will" sounds like a very continuous concept, doesn't it? //It might look like intelligence isn highly discontinuous if you examine humans too closely :)

(errata)

[TheCouchPotatoFamine]

took too long to write and i didn't read it again, sorry about the typos at the bottom and the forgotten parallel about what a digital delay pedal entails. The difference between iron-tape and RAM. It's not the best example. ah well.

Re:(errata)

[fyngyrz]

The error in your idea is that a serial computer has absolutely zero problem 100% accurately modeling a parallel computer; a real parallel machine is probably much faster (that's the point, of course, so you'd hope so) but it won't get any different answer to a problem than the serial computer modeling the parallel one. That's what is being said here - there is no indication that there are any local, small systems of any kind that cannot be modeled on a serial computer; hence, there is no scientific reason to postulate that intelligence represents such an impossible-to-model system. Such a claim (intelligence being impossible to model) is simply an indulgence in hubris until or unless we know a lot more about the problem(s.)

Mod Up

[TheCouchPotatoFamine]

mod the godel gotchas gabbed garishly up

Visual Search?

[Chas]

Mother: Go find your little brother
Older Brother: Found him! He's behind the sofa.

*RING!*

Mother: Hello?
Voice On The Other End Of The Line: Ma'am, this is Pubert Skewya. I'm a lawyer for Duey, Cheatham, and Howe. We represent Hitachi.
Mother: Uhm. Yeah? So what?
VOTOEOTL: Ma'am, we have a record that you just encouraged your son to violate our client's patent on visual searches. Natually, we'll settle out of court for one billion dollars, American. If you refuse, with the state of the economy as it is, we'll go after you in court, but we'll go after you for one billion dollars, Canadian. If you act now, and concede to our extor*COUGH*rightful demands, you'll save yourself money in the long run.
Mother: Uhm. Yeah. Who is this really?
VOTOEOTL: Ma'am, this is serious. Our client has a patent on visual searches. Every time you tell your son to go look for something, you're contributing to the violation of our client's patents.
Mother: And I know ONE young man who's going to get his ass beaten for putting one of his idiot friends up to this stupid little prank...
*CLICK*

VOTOEOTL to the rest of his call center: SHIT! That's like the millionth one!

What about Scale-Invariant techniques?

[Ayanami Rei]

http://web.engr.oregonstate.edu/~hess/index.html#% 5B%5BSIFT%20Feature%20Detector%5D%5D

These techniques allow you to preprocess the image into a set of feature vectors which can be organized into a database and indexed with some effectiveness.

Airplane Porn?

[Ayanami Rei]

http://forums.somethingawful.com/showthread.php?th readid=2551167

NWS http://ft.mirror.waffleimages.com/files/e1/e165187 4770c2c2a42433b4636f3c9621942c7c3.png

actually, you outlined my point in big O terms...:

[TheCouchPotatoFamine]

I agree that a serial system can emulate a parallel one. You can even determine how much longer a serial computer will take to execute a massively parallel program. Problem is, the time it takes to solve a problem scales very very badly with the complexity of the problem. If you have a true parallel machine (as possible using physical phenomena such that analog systems use in one form or another) and not just a system that computes in parallel, but in parallel of thousands or hundreds of thousands of nodes in just a few "steps" - you can't really always define the processing of a real neural network as "step based".

That's my argument, that there is an entirely different way of solving problems in a discrete "digital" system then a continuous "analog" system. They have their strengths and *compliment* each other, not *equal* each other, in terms of their strengths.

Trying to model certain aspects of intelligence in silicon is like trying to make a rocket out of some number of aerosol deodorant cans - theoretically you can fly if you just have enough but in practice you'll always need more aerosol then space available on the thing to be lifted! And it never gets better using aerosol - lifting a person is just as hard as lifting a space-shuttle.

On certain intelligence related problems, digital will *never* enjoy the economies of scale that analog neural network design provides. This is a deep intuition after much study. Bet on it.

decades old

[oohshiny]

The approach is decades old... but no doubt, it's newly patented.

Speed issues

[Bob of Dole]

I wonder if they've managed to solve the slowness of this sort of search anyway other than just throwing a lot of boxes at it?
My own system does 13 million images in about a minute, but with enough RAM to fit the dataset in memory I can do 10-20 seconds.
I hope they're not just using a cluster to speed up access, that's a workable solution but it doesn't really help those of us who can't afford a dozen boxes to power their searcher.

Great!

[Anonymous Coward]

So it can recognize shapes and colour distribution?
Then it can auto-categorize all my pr0n ;-)))

Already done

[12357bd]

To compare million images in a second on a comodity pc was already done since 2004/2005 see http://www.immem.com/en/. By now the state of the art is to compare a video stream against a 24h video pool in realtime, using this technology.

Oh boy!

[AndyChrist]

So now maybe I won't have to go to 4chan to find moar of some chick whose name I don't know.

The human brain does one thing only

[master_p]

The human brain does one thing only: pattern matching.

More specifically, the body sensors ask questions to the brain, and the brain searches its database of experiences to find the experience which maximizes survival in the current situation. Once the experience is found, it is activated and answers are sent to the sensors.

The above mechanism has been developed because mathematical logic can not prove that a situation is dangerous for an animal or not. For example, it can not be proved that facing a lion is dangerous, because not all the facts about the environment and the lion's status are known. But pattern matching can 'prove' that something is dangerous by recalling past experiences or knowledge.

That is the reason we have religions: we could not understand certain physical phenomena around us, so we had to invent a reason for them...and since we were not able to reproduce those phenomena, someone with higher capabilities than us must have been responsible for those phenomena. For example, when we saw thunder, we did not understand how they were produced, and since we did not produce them, we had to believe that someone else did them, someone with higher powers than us.

Needing to comprehend those phenomena was crucial to our survival: by "understanding" that a god did not make thunders unless we disobeyed his rules, we could keep our dopamine levels down, and thus being calm and be able to assess the various dangers better.

In conclusion, it's all about maximizing survival. True AI will come only when the above mechanism will be transfered to mechanical devices. By AI, it does not mean machines will suddenly write poems, but that machines would not need to be programmed but taught.

Re:HDD

[Anonymous Coward]

Yes, I'm an idiot. Spoke before I thought.

But come on now, flamebait? Hardly. Offtopic, overrated, wrong, whatever. But hardly flamebait.

Lame.

Naive Implementation satisfying 92% of queries...

[Anonymous Coward]

if ($search_request =~ /.*/){
        return images_matching_query
                ("select img from images where lower(name) like '%paris%hilton%');
}

Re:actually, you outlined my point in big O terms.

[fyngyrz]

On certain intelligence related problems, digital will *never* enjoy the economies of scale that analog neural network design provides. This is a deep intuition after much study. Bet on it.

Oh, I do. What I don't count on is that NN's, or more generally, vague models of animal function, are the only, or even the, answer. I have said all along that the general purpose computer can model anything at little or no extra hardware cost; once - if - we find what works, by all means, hand it to the engineers and let them take the most active elements and create hardware specialized to do those operations. I can just about guarantee that this is how it'll go. The problem with (for instance) hardware neural net development is that creating non-NN elements requires painstaking hardware design; in a computer, you just model the inputs, engine, and outputs, and go on with your day. This is more generally true for any hardware-based approach. Even if it is the right path, it's a slow path. Simulate, identify, then go to hardware as and if required. That's what makes the most sense.

Photosynth

[thoughtlover]

This looks similar to how Photosynth stores and makes image correlation. And I'd say, it's one of the more impressive things I've seen Microsoft do, but I think they bought this technology.

http://www.youtube.com/watch?v=s-DqZ8jAmv0

Re:Desktop search with image pattern recognition A

[smellotron]

Since camera makers dropped the ball and don't have easy intuitive image tagging capability built into the camera.

I wouldn't say camera makers have dropped the ball... I'd say you're just looking at the wrong ball. Canon, Nikon et al are coming out with fantastic low-end DSLRs with "easy intuitive" controls for ISO, aperture, shutter, white balance... things that are important for taking pictures. Adding tagging directly to the camera would either mean a clunky typing interface (not enough buttons), an expensive and fragile touchscreen, or an expensive (both $$$ and electricity) speech recognition chip.

I'm happy with a camera that does a really good job at taking pictures. Leave the tagging for some software "image productivity" suite.

Glad this is public now

[BananaStewGuy]

I got a chance to see this software in Japan the last time I visited Hitachi's Central Research Labs. It was impressive. Unfortunately, I couldn't tell anyone about it because it was still under wraps. Now that it's out in the open, here's a post with some details. Briefly, it does rely on pre-indexing of the images, it doesn't rely on any text tagging, and it's not intended to compete with Google Image Search et. al. It is intended as an Enterprise application. It is remarkably good at finding faces, even when you don't tell it you're looking for faces. And it even works on video clips. Unfortunately, they didn't give me a copy to take home.