It uses a (cheap) commodity digital camera, combined with a smart tripod, good photo stitching software, and a nice Flash UI to give you highly zoomable panoramas. The CMU thing has been around for a while --- over a year at least, plus I'm pretty sure you can get one of the tripod mounts if you participate in the beta and create your own.
I haven't been able to try pipes yet, because the site is down. But I have a colleague who built something that sounds quite similar. It's called the Javascript Device Architecture. At the link, you can find demos and downloadable code. So, if you can't get to pipes, you could always try JDA out;-)
I didn't build it all myself, I was just on the team, but we made a rig that simulated the Matrix BulletTime effect using 32 $20 Mattel Barbie digital cameras. The cameras were mounted on foam core and corrugated plastic, arranged in a big circle, and we used truck mirrors to get a wide angle effect. All the cameras were wired to a central triggering circuit, and we used a garage door opener as a remote control. You would go into the center of the circle with your friends and some props, do something crazy, and hit the remote. The cameras would fire --- then all the pictures would be sent to a printer which would print out a flip-book on cardstock so that you could see a low-tech animation of yourself spinning around doing whatever.
Actually, I did read that part of the article. I don't pretend to fully understand the details, but it seems like that addresses reactor malfunctions, not the risk of nuclear material being scattered.
Other people have helpfully pointed out that if you use the engine only in space (and possibly even assemble it there), you can work really hard and secure the nuclear material on the way up to orbit. Others have also pointed out that we certainly have put nuclear materials in orbit before (though that's not really an iron-clad argument that it's a safe thing, just that we haven't fucked up yet).
This isn't meant as a panicky "omg! nucular!" question. But we have seen a few space craft blow up spectacularly. Now, I assume the designers are bright enough that these engines could not actually produce a nuclear explosion, but wouldn't a conventional explosion at high altitude run a high risk of scattering nuclear material all over the place? Is there a good reason I shouldn't be worried about that?
David Gelernter built a system called Lifestreams that basically claimed that time-ordered series plus some simple search and organization operators was everything you needed. It always seemed like a pretty good idea to me.
That said, if you do have metadata available, you can do a lot with it.
Actually, we did that a couple of years ago using Mattel "Barbie" digital cameras. They were being sold below cost for $20 or something. It was quite a bitch to do the custom circuit to get them to all fire simultaneously and then download images in parallel. We directed the little "movie" to a printer that printed on perforated cardstock paper, so you could make a flipbook of your little "matrix" effect.
That was a hack for a big party a client was having. Later on we did the same thing using more reliable hardware with better resolution (and USB: always nice) for the exhibit/tradeshow industry. You can rent one here:
http://www.flip360.com/
But yeah, I expect cheap digital cameras will make more and more of these lo-fi real-time special effects things possible.
An actually useful location-based service
on
Real-World Hyperlinks
·
· Score: 4, Informative
If you're using a computer at CMU (or one of Telerama's wireless hotspots in Pittsburgh), you can find out when the next bus comes near you at bus.maya.com.
Perhaps it's not as glamorous as streaming Quicktime movies to your phone, but it's probably more useful;-)
That said, I hope someone solves the location-based services infrastructure problem. The bus hack depends on mapping IP addresses to lat/lons, which is incredibly brittle and evil.
I think MIT developed something along these lines a long time ago. (Here's a link.) The idea was not to empower the government, but to provide a sort of Super PDA for the individual. Oddly enough, I think it uses Emacs.
Of course, half the world seems to be blogging all the time anyway, which tend to be weak on the indexing and searching, but provide a nice low barrier-to-entry for inputting all kinds of trivial crap about one's life.
It's not necessarily entirely about dystopian government power;-)
A corollary is "an immediate consequence of a result already proved. Corollaries usually state more complicated theorems in a language simpler to use and apply."
You're proposing, I don't know, another axiom or something.
I'm pretty sure in Distraction random untrained people are always helping out on building sites because the pieces of the building know how to put the building together, and helpfully direct you.
Kind of cool that someone is doing it, even if at a small scale.
For the record (I'm one of the researchers on the project), the prototype can control shelf-top mp3 players (we've implemented the control protocol for the AudioRequest device). (And one of the paper prototypes considered an office phone and voicemail system, though we haven't implemented that, so we're pretty confident we could handle your answering machine with no problem.)
The system will control anything you write a specification for.
Some neat features that the article sort of glossed over: the system generates UIs that are demonstrably better than traditional remote controls. It requires no programming at all. It can generate multi-modal interfaces (combining speech and GUI). It can generate interfaces for any form factor remote control device (this is the same flexibility that allows multi-modal interfaces).
The CMU page has lots of information about what the system is really capable of.
Oh, and yeah, the system is currently implemented in Java and it runs on Linux as well as PocketPC (in fact, the widgets look better on the Zaurus).
Unfortunately, our system is not smart enough to read slashdot and detect duplicate posts.
(Bias alert: I was the project manager for MAYA on the PUC.)
The article is a little light on the details. The PUC does some things that systems like the Pronto can't.
First, there is NO programming. Zero, nada. The appliance sends an abstract specification of its features to the remote, which then automatically generates an appropriate interface. This interface can be tailored to the needs or preferences of the user, and can take advantage of idiosyncratic features of the device being used as a controller (for instance, we could use a jog-wheel for volume if one is available).
The interface specification language is VERY abstract. So abstract that we can automatically generate interfaces in multiple modalities, like, say speech. We're working on mixing modalities.
Plus, our remote is two way. If someone else changes the channel, the remote's display reflects the change. This lets you design much better interfaces because you can hide features that aren't currently available. This minimizes the number of buttons you need and drastically reduces mode errors.
P.S. The whole needing-a-laptop thing is because we hacked all this into a $75 shelf stereo we got from Best Buy. Obviously a production system would not have any such kludge.
Actually, the prototype client is Java. Runs fine on the Sharp Zaurus (and the widgets look nicer).
Anyway, the important facet of the system is the XML specification and protocol, which could be implemented on any platform. We'd love to do a cellphone...
I'm the Mike Higgins quoted in the article. The article is less than informative. If you want better info, try CMU's page. MAYA has a page too but it's a bit more markety.
The system is actually pretty neat: no programming is required. We're working on generating multi-modal interfaces. The remote is two-way, so it gives much better feedback than traditional remotes (leading to better interfaces).
My company does some "pervasive computing" research. In many situations it's convenient to use an iPAQ because they're quite powerful, even though the UI sucks. (Occasionally we've even done prototypes by hiding an iPAQ running Linux *inside* another piece of plastic.)
We've found that the Zaurus actually runs Java better than the iPAQ (the widgets come out nicer and it seems to suffer fewer drawing bugs). I also think the UI is marginally nicer than Pocket PC.
But these are all just in situations where we need a lot of horsepower and a really nice screen. For day to day use a Palm gets the job done best.
P.S. If you're curious, here are some pages about two projects that we've used iPAQs for:
Though I don't have the text in front of me, I believe that describing computers as being good at "fetishistically counting things" was used by Neal Stephenson in Snow Crash to describe the operation of the Metaverse. In particular, he describes the fetishistic attachment we all have for 2 and powers of 2.
Yep. The Pittsburgh office just got cut (taking friends of mine with it). This doubtless means that the CMU edition of Software Engineering for Web Applications is going to go too.
It sounds like they're becoming just another totally corporate Web house. One hears that the entire code base is going to Java, not because of its technical superiority (servlets are immature compared to most other server-side dev technologies) but because it's easy to sell Java to big enterprise businesses.
With Greenspun isolated by the Board of Directors I think a lot of the soul of the company is gone.
I have one too. Compared to any Palm device the apps are poor and the speed is TERRIBLE.
Frankly, I found it painful to use.
So I took it apart. I'm thinking of nailing it to a wall and using it to run some sort of dedicated app. Maybe recipes in the kitchen or an mp3-control panel.
Sounds an awful lot like feeble human pattern recognition to me. They are both created by the same species to make sense of its problems (abstract, applied, practical, or unpractical), whether they look inward or outward in their respective domains. If you think about the actual agents creating those bodies of knowledge they probably aren't so far apart. Both physics and mathematics have made contributions to each other and provided new direction and insight from new theory and discovery. Haven't you read your Kuhn?
Well, I think this is a pretty decent candidate for an explanation. But it basically rejects the "special character" of mathematics. Math becomes a feature of human psychology and our *perception* of the universe, not a fundamental feature of the universe itself. And the "truths" we discover are maybe just local phenomena.
Many folks find this an upsetting viewpoint, and it does tend to devour itself. (If we're just "recognizing patterns" isn't there something we can say about pattern-recognition in general? Wouldn't that be a transcendant, mathematical claim? Or, again, are we just lucky that pattern-recognition happens to work for the moment?)
"What do you mean by mathematical truths? If you mean theorems, aren't the theorems provable as consequences of particular assumed axioms? If you mean the axioms themselves, well, then mathematics isn't an especially special case is it? Isn't any system going to have to have certain fundamental axioms you take as true, that aren't proved?"
Oh, but this is the whole point of the incompleteness result. There are truths, that is statements in the formal language, which are not provable, that is, not theorems. They are *just true* and you can't prove them. This is different from being an axiom as well. (Axioms you can think of as being their own degenerate proofs.)
And it doesn't help matters to simply identify the unprovable truths as axioms: this just generates more unprovable truths. The incompleteness result is very robust.
Technically, if you want to know how a mathematical truth is defined, you need to study model theory and formal logic. A model generates a semantics (and truth values) for a given formal system. Otherwise the formal system is just symbols. You can think of the model as a mapping from well-formed formulas in the formal language to truth values (typically true and false). The inference rules used to manipulate the formal system will be truth-preserving in the model, etc. The incompleteness result is telling us that there are necessarily a bunch of wffs that are mapped to "true" in the model that we can never reach using our inference rules (our means of proof). And this result obtains under a very wide range of circumstances. (You aren't going to get away from it by messing with the model, or the axioms, or the inference rules, without destroying the "interesting" features of the formal system.)
But this is all background and doesn't speak directly to Chaitin's results.
I happened to sit in on the lecture at CMU. Certainly Chaitin's results do owe a lot to Godel and Turing, but it's not just a rehash.
Here's what's interesting to me:
First, it's mysterious that mathematical truths are applicable to the "real world". This is a philosophical question that people have struggled over for a long time. Why is it that abstract mathematical structures discovered without any reference to physics often later turn out to be useful in physical theories?
Now consider what Chaitin is saying. Very few mathematical truths have any structure at all. That means that we can't prove the vast majority of true theorems, and if you were to pick a mathematical truth out of the air it is unlikely in the extreme that you could find a proof for it.
Put these two facts together. Isn't it awfully surprising that mathematics is so successful at describing the real world? Math is full of unproveable truths, and yet, we seem to be able to prove a bunch of really useful things.
Now why should that be? I don't know.
If you're an optismist, you might say, how lucky! It's a good thing that the universe is structured in a way that's mostly congruent with the proveable sections of mathematics.
If you're a pessimist, you might wonder how long our luck is going to last.
Slashdot Mirror
The folks at CMU have a similar thing:
http://gigapan.org/
It uses a (cheap) commodity digital camera, combined with a smart tripod, good photo stitching software, and a nice Flash UI to give you highly zoomable panoramas. The CMU thing has been around for a while --- over a year at least, plus I'm pretty sure you can get one of the tripod mounts if you participate in the beta and create your own.
Slate has a pretty decent write-up about the bill.
I haven't been able to try pipes yet, because the site is down. But I have a colleague who built something that sounds quite similar. It's called the Javascript Device Architecture. At the link, you can find demos and downloadable code. So, if you can't get to pipes, you could always try JDA out ;-)
I didn't build it all myself, I was just on the team, but we made a rig that simulated the Matrix BulletTime effect using 32 $20 Mattel Barbie digital cameras. The cameras were mounted on foam core and corrugated plastic, arranged in a big circle, and we used truck mirrors to get a wide angle effect. All the cameras were wired to a central triggering circuit, and we used a garage door opener as a remote control. You would go into the center of the circle with your friends and some props, do something crazy, and hit the remote. The cameras would fire --- then all the pictures would be sent to a printer which would print out a flip-book on cardstock so that you could see a low-tech animation of yourself spinning around doing whatever.
Here's some propaganda about the project.
Actually, I did read that part of the article. I don't pretend to fully understand the details, but it seems like that addresses reactor malfunctions, not the risk of nuclear material being scattered.
Other people have helpfully pointed out that if you use the engine only in space (and possibly even assemble it there), you can work really hard and secure the nuclear material on the way up to orbit. Others have also pointed out that we certainly have put nuclear materials in orbit before (though that's not really an iron-clad argument that it's a safe thing, just that we haven't fucked up yet).
This isn't meant as a panicky "omg! nucular!" question. But we have seen a few space craft blow up spectacularly. Now, I assume the designers are bright enough that these engines could not actually produce a nuclear explosion, but wouldn't a conventional explosion at high altitude run a high risk of scattering nuclear material all over the place? Is there a good reason I shouldn't be worried about that?
David Gelernter built a system called Lifestreams that basically claimed that time-ordered series plus some simple search and organization operators was everything you needed. It always seemed like a pretty good idea to me.
That said, if you do have metadata available, you can do a lot with it.
I was very confused and kept going to this Mike Rowe guy's site instead. I tell ya. There oughta be a law.
Actually, we did that a couple of years ago using Mattel "Barbie" digital cameras. They were being sold below cost for $20 or something. It was quite a bitch to do the custom circuit to get them to all fire simultaneously and then download images in parallel. We directed the little "movie" to a printer that printed on perforated cardstock paper, so you could make a flipbook of your little "matrix" effect.
t _f ilmmakers_360.mtml
More info here:
http://www.maya.com/web/what/clients/what_clien
That was a hack for a big party a client was having. Later on we did the same thing using more reliable hardware with better resolution (and USB: always nice) for the exhibit/tradeshow industry. You can rent one here:
http://www.flip360.com/
But yeah, I expect cheap digital cameras will make more and more of these lo-fi real-time special effects things possible.
If you're using a computer at CMU (or one of Telerama's wireless hotspots in Pittsburgh), you can find out when the next bus comes near you at bus.maya.com. Perhaps it's not as glamorous as streaming Quicktime movies to your phone, but it's probably more useful ;-)
That said, I hope someone solves the location-based services infrastructure problem. The bus hack depends on mapping IP addresses to lat/lons, which is incredibly brittle and evil.
Perhaps Sony was just experimenting with anti-gravity... I think they're going to need to get the telcos to use higher voltage, though.
I think MIT developed something along these lines a long time ago. (Here's a link.) The idea was not to empower the government, but to provide a sort of Super PDA for the individual. Oddly enough, I think it uses Emacs.
;-)
Another interesting system was Gelernter's LifeStreams, which time-indexed everything...
Of course, half the world seems to be blogging all the time anyway, which tend to be weak on the indexing and searching, but provide a nice low barrier-to-entry for inputting all kinds of trivial crap about one's life.
It's not necessarily entirely about dystopian government power
Throwing karma to the wind... that's not a corollary.
From Mathworld:
A corollary is "an immediate consequence of a result already proved. Corollaries usually state more complicated theorems in a language simpler to use and apply."
You're proposing, I don't know, another axiom or something.
(All right, I apologize. I'm a geek.)
In any case, it's probably correct.
I'm pretty sure in Distraction random untrained people are always helping out on building sites because the pieces of the building know how to put the building together, and helpfully direct you.
Kind of cool that someone is doing it, even if at a small scale.
For the record (I'm one of the researchers on the project), the prototype can control shelf-top mp3 players (we've implemented the control protocol for the AudioRequest device). (And one of the paper prototypes considered an office phone and voicemail system, though we haven't implemented that, so we're pretty confident we could handle your answering machine with no problem.)
The system will control anything you write a specification for.
Some neat features that the article sort of glossed over: the system generates UIs that are demonstrably better than traditional remote controls. It requires no programming at all. It can generate multi-modal interfaces (combining speech and GUI). It can generate interfaces for any form factor remote control device (this is the same flexibility that allows multi-modal interfaces).
The CMU page has lots of information about what the system is really capable of.
Oh, and yeah, the system is currently implemented in Java and it runs on Linux as well as PocketPC (in fact, the widgets look better on the Zaurus).
Unfortunately, our system is not smart enough to read slashdot and detect duplicate posts.
(Bias alert: I was the project manager for MAYA on the PUC.)
The article is a little light on the details. The PUC does some things that systems like the Pronto can't.
First, there is NO programming. Zero, nada. The appliance sends an abstract specification of its features to the remote, which then automatically generates an appropriate interface. This interface can be tailored to the needs or preferences of the user, and can take advantage of idiosyncratic features of the device being used as a controller (for instance, we could use a jog-wheel for volume if one is available).
The interface specification language is VERY abstract. So abstract that we can automatically generate interfaces in multiple modalities, like, say speech. We're working on mixing modalities.
Plus, our remote is two way. If someone else changes the channel, the remote's display reflects the change. This lets you design much better interfaces because you can hide features that aren't currently available. This minimizes the number of buttons you need and drastically reduces mode errors.
Find out more. Check out CMU's page and MAYA's more markety page.
P.S. The whole needing-a-laptop thing is because we hacked all this into a $75 shelf stereo we got from Best Buy. Obviously a production system would not have any such kludge.
Actually, the prototype client is Java. Runs fine on the Sharp Zaurus (and the widgets look nicer).
Anyway, the important facet of the system is the XML specification and protocol, which could be implemented on any platform. We'd love to do a cellphone...
I'm the Mike Higgins quoted in the article. The article is less than informative. If you want better info, try CMU's page. MAYA has a page too but it's a bit more markety.
The system is actually pretty neat: no programming is required. We're working on generating multi-modal interfaces. The remote is two-way, so it gives much better feedback than traditional remotes (leading to better interfaces).
My company does some "pervasive computing" research. In many situations it's convenient to use an iPAQ because they're quite powerful, even though the UI sucks. (Occasionally we've even done prototypes by hiding an iPAQ running Linux *inside* another piece of plastic.)
We've found that the Zaurus actually runs Java better than the iPAQ (the widgets come out nicer and it seems to suffer fewer drawing bugs). I also think the UI is marginally nicer than Pocket PC.
But these are all just in situations where we need a lot of horsepower and a really nice screen. For day to day use a Palm gets the job done best.
P.S. If you're curious, here are some pages about two projects that we've used iPAQs for:
Personal Information Portal (very out of date)
Personal Universal Controller (with CMU) We've tried the Zaurus on the latter and it works much better.
Though I don't have the text in front of me, I believe that describing computers as being good at "fetishistically counting things" was used by Neal Stephenson in Snow Crash to describe the operation of the Metaverse. In particular, he describes the fetishistic attachment we all have for 2 and powers of 2.
Mike
Yep. The Pittsburgh office just got cut (taking friends of mine with it). This doubtless means that the CMU edition of Software Engineering for Web Applications is going to go too.
It sounds like they're becoming just another totally corporate Web house. One hears that the entire code base is going to Java, not because of its technical superiority (servlets are immature compared to most other server-side dev technologies) but because it's easy to sell Java to big enterprise businesses.
With Greenspun isolated by the Board of Directors I think a lot of the soul of the company is gone.
I have one too. Compared to any Palm device the apps are poor and the speed is TERRIBLE.
Frankly, I found it painful to use.
So I took it apart. I'm thinking of nailing it to a wall and using it to run some sort of dedicated app. Maybe recipes in the kitchen or an mp3-control panel.
I would recommend saving your money.
Anonymous Coward said:
Well, I think this is a pretty decent candidate for an explanation. But it basically rejects the "special character" of mathematics. Math becomes a feature of human psychology and our *perception* of the universe, not a fundamental feature of the universe itself. And the "truths" we discover are maybe just local phenomena.
Many folks find this an upsetting viewpoint, and it does tend to devour itself. (If we're just "recognizing patterns" isn't there something we can say about pattern-recognition in general? Wouldn't that be a transcendant, mathematical claim? Or, again, are we just lucky that pattern-recognition happens to work for the moment?)
pangloss said:
Oh, but this is the whole point of the incompleteness result. There are truths, that is statements in the formal language, which are not provable, that is, not theorems. They are *just true* and you can't prove them. This is different from being an axiom as well. (Axioms you can think of as being their own degenerate proofs.)
And it doesn't help matters to simply identify the unprovable truths as axioms: this just generates more unprovable truths. The incompleteness result is very robust.
Technically, if you want to know how a mathematical truth is defined, you need to study model theory and formal logic. A model generates a semantics (and truth values) for a given formal system. Otherwise the formal system is just symbols. You can think of the model as a mapping from well-formed formulas in the formal language to truth values (typically true and false). The inference rules used to manipulate the formal system will be truth-preserving in the model, etc. The incompleteness result is telling us that there are necessarily a bunch of wffs that are mapped to "true" in the model that we can never reach using our inference rules (our means of proof). And this result obtains under a very wide range of circumstances. (You aren't going to get away from it by messing with the model, or the axioms, or the inference rules, without destroying the "interesting" features of the formal system.)
But this is all background and doesn't speak directly to Chaitin's results.
Here's what's interesting to me:
First, it's mysterious that mathematical truths are applicable to the "real world". This is a philosophical question that people have struggled over for a long time. Why is it that abstract mathematical structures discovered without any reference to physics often later turn out to be useful in physical theories?
Now consider what Chaitin is saying. Very few mathematical truths have any structure at all. That means that we can't prove the vast majority of true theorems, and if you were to pick a mathematical truth out of the air it is unlikely in the extreme that you could find a proof for it.
Put these two facts together. Isn't it awfully surprising that mathematics is so successful at describing the real world? Math is full of unproveable truths, and yet, we seem to be able to prove a bunch of really useful things.
Now why should that be? I don't know.
If you're an optismist, you might say, how lucky! It's a good thing that the universe is structured in a way that's mostly congruent with the proveable sections of mathematics.
If you're a pessimist, you might wonder how long our luck is going to last.