I get the feeling that all the other posters pointing out that firefox already supports "highlight and search" don't get the point. I think the novelty here is that you can highlight huge sections of text, send it to Y!Q, and it will find other pages that are related, given that context.
After playing with it, I can see some how Y!Q is different - I went to a random article on Google news, highlighted an entire paragraph of text, and sent it to Y!Q. It returned a bunch of pages related to that original news article. When you do the same thing with Google search, not surprisingly, it only returns one page.
So, yes, Y!Q does something different. I don't know if it's entirely useful, but it's not entirely redundant, either.
Both Apples and Oranges are platforms for seed distribution. Both are fruits. Both make it easier for people to live (by providing vitamins, minerals, etc.)
It's hard to set odds on something like this, but the most informed I've seen would give us about even odds of having a populated area smashed up (damage as much as a trillion dollars) sometime in the next millenium.
A 50% chance of a trillion dollars of damage over a thousand years means it'd be worth it to spend up to a half a billion dollars a year to prevent an asteroid smash!:)
This mod to a Pelican Light Shield Advance was pretty easy to do and the parts are readily available. I'm pretty happy with it - it's better than anything I've seen for sale.
The portablemonopoly solution is better, though - if for no other reason than eliminating the glare from the light source.
Well, I was just talking about mangling all images so that steganographic content was no longer recoverable, not detection nor decoding of it - you'd be altering images which had no hidden data, too, but it shouldn't be noticeable. You do bring up an interesting point about steganography which can withstand lossy conversions such as printing and rescanning.
I would guess that a method which could withstand such manipulations would have a lower bandwidth -- there'd have to be some kind of error correction/redundancy in those methods. I've heard of watermarking schemes which are extremely resilient, but a watermark contains considerably less information than a steganographically encoded image.
I'm no expert in information theory, but shouldn't there be a way to measure how much information would be able to survive a filter? Would it be enough to prevent hiding of reconnaisance photos and maps?
Speaking of online media (in another sense) - I've heard reports in the past that claimed that Bin Laden steganographically hid data inside of images and postings on public message boards to communicate.
I was wondering how feasible it would be to create filters for Apache and/or IIS which would strip images of potential steganographic information. For instance, GIF images could be converted to JPG's and then back to GIF's before being served - this would probably eliminate hiding data in the LSB. (Obviously you'd have to do some caching) You could do similar things to other media, like WAV and MP3 files.
How hard is this? Could it be used to prevent use of public servers as repositories for steganographic media? Would it work? Would it be used widely enough to be effective?
Metropolis and simulated annealing are mostly the same, I think. It basically works like this:
1) Start off with a proposed solution (I suppose the "molecule").
2) Evaluate its fitness, let's call that F. The fitness of the keypad was probably the weighted sum of stylus transition times from letter to letter. Lower fitness would be better (I know it doesn't sound right, but it's used by a lot of Evo-algorithm researchers).
3) Randomly mutate the "molecule" to something similar to the original. Here, the researchers probably randomly swapped key positions, or "atoms".
4) Evaluate the fitness of the new "molecule".
5) Calculate the change in fitness, let's call that dF.
6) If dF < 0, then use the new "molecule", and go back to step 2.
7) Calculate the value of p=e^(-dF/T), where T is a constant. This value p is the probability you will switch to the new "molecule".
8) Go back to step 2.
I think the difference between Metropolis and simulated annealing is that T, the temperature of the system, goes down with time in simulated annealing.
Anyhow, the reason I quoted the words "molecule" and "atom" is because they are only loosely related to the algorithm. And to call this thing a "molecular" keyboard is misleading (well, except for the fact that most matter here on earth is molecular). I could go ahead and replace the words "molecule" with "tinkertoy" and "atom" with "rods and wheels" and I'll sell you a tinkertoy keyboard. It could go well with that tinkertoy computer they built at MIT...
I agree - I don't quite see the connection to optical computing here (but maybe someone who does could enlighten us?).
However, I bet this lens could be useful in optical networks - perhaps in gathering and converting light back into electric signals. If I recall correctly, the ends of fibers are shaped into lenses to focus the light onto the photo-diode. Increasing the efficiency of that lens could extend the usable length of a fiber.
Some people have suggested that software could be used to perform the
interpolation on extracted music, stripping out the bits that the record
companies added in. The trouble with this approach is that, once the data has
been extracted, the CIRC encoding is no longer visible. It may not be easy to
tell where the glitches are. For example, it should be possible to create a
low-level but rhythmic distortion that will be noticeable, annoying, and
difficult to identify automatically.
Here's my guess as to why it's harder than just "fixing it in software.":
It sounds like the glitches are being put into the error correction layer protecting the data on the CD. When you read data off of a CD-ROM, the error-correction bits are used to check the integrity of the data, and if a problem is found, it's fixed. All you get is the "corrected data".
While reading audio off of a CD, the decoder uses the bits only as error-detection, and whenever it detects a glitch, it just throws it away - something you wouldn't want to do with data.
In any case, it sounds like by the time you get the data from a CD-ROM drive, those error correction bits are long gone, and detecting that data has been intentionally corrupted is far from trivial.
Now, I may be wrong on all of this - if it is possible to read the "CIRC encoding" from the CD, then it should be fixable in software... But from the sounds of it, this may have to be fixed in hardware.
Slashdot Mirror
I get the feeling that all the other posters pointing out that firefox already supports "highlight and search" don't get the point. I think the novelty here is that you can highlight huge sections of text, send it to Y!Q, and it will find other pages that are related, given that context.
After playing with it, I can see some how Y!Q is different - I went to a random article on Google news, highlighted an entire paragraph of text, and sent it to Y!Q. It returned a bunch of pages related to that original news article. When you do the same thing with Google search, not surprisingly, it only returns one page.
So, yes, Y!Q does something different. I don't know if it's entirely useful, but it's not entirely redundant, either.
Both Apples and Oranges are platforms for seed distribution. Both are fruits. Both make it easier for people to live (by providing vitamins, minerals, etc.)
A 50% chance of a trillion dollars of damage over a thousand years means it'd be worth it to spend up to a half a billion dollars a year to prevent an asteroid smash!
This mod to a Pelican Light Shield Advance was pretty easy to do and the parts are readily available. I'm pretty happy with it - it's better than anything I've seen for sale.
The portablemonopoly solution is better, though - if for no other reason than eliminating the glare from the light source.
Well, I was just talking about mangling all images so that steganographic content was no longer recoverable, not detection nor decoding of it - you'd be altering images which had no hidden data, too, but it shouldn't be noticeable. You do bring up an interesting point about steganography which can withstand lossy conversions such as printing and rescanning.
I would guess that a method which could withstand such manipulations would have a lower bandwidth -- there'd have to be some kind of error correction/redundancy in those methods. I've heard of watermarking schemes which are extremely resilient, but a watermark contains considerably less information than a steganographically encoded image.
I'm no expert in information theory, but shouldn't there be a way to measure how much information would be able to survive a filter? Would it be enough to prevent hiding of reconnaisance photos and maps?
Speaking of online media (in another sense) - I've heard reports in the past that claimed that Bin Laden steganographically hid data inside of images and postings on public message boards to communicate.
I was wondering how feasible it would be to create filters for Apache and/or IIS which would strip images of potential steganographic information. For instance, GIF images could be converted to JPG's and then back to GIF's before being served - this would probably eliminate hiding data in the LSB. (Obviously you'd have to do some caching) You could do similar things to other media, like WAV and MP3 files.
How hard is this? Could it be used to prevent use of public servers as repositories for steganographic media? Would it work? Would it be used widely enough to be effective?
You mean I'm going to have to wait 3 and a half months before I can get my imports?
Metropolis and simulated annealing are mostly the same, I think. It basically works like this:
1) Start off with a proposed solution (I suppose the "molecule").
2) Evaluate its fitness, let's call that F. The fitness of the keypad was probably the weighted sum of stylus transition times from letter to letter. Lower fitness would be better (I know it doesn't sound right, but it's used by a lot of Evo-algorithm researchers).
3) Randomly mutate the "molecule" to something similar to the original. Here, the researchers probably randomly swapped key positions, or "atoms".
4) Evaluate the fitness of the new "molecule".
5) Calculate the change in fitness, let's call that dF.
6) If dF < 0, then use the new "molecule", and go back to step 2.
7) Calculate the value of p=e^(-dF/T), where T is a constant. This value p is the probability you will switch to the new "molecule".
8) Go back to step 2.
I think the difference between Metropolis and simulated annealing is that T, the temperature of the system, goes down with time in simulated annealing.
Anyhow, the reason I quoted the words "molecule" and "atom" is because they are only loosely related to the algorithm. And to call this thing a "molecular" keyboard is misleading (well, except for the fact that most matter here on earth is molecular). I could go ahead and replace the words "molecule" with "tinkertoy" and "atom" with "rods and wheels" and I'll sell you a tinkertoy keyboard. It could go well with that tinkertoy computer they built at MIT...
Hi! How are you? I send you this MP3 in order to have your listening. See you later. Thanks.
I agree - I don't quite see the connection to optical computing here (but maybe someone who does could enlighten us?).
However, I bet this lens could be useful in optical networks - perhaps in gathering and converting light back into electric signals. If I recall correctly, the ends of fibers are shaped into lenses to focus the light onto the photo-diode. Increasing the efficiency of that lens could extend the usable length of a fiber.
From the article:
Some people have suggested that software could be used to perform the
interpolation on extracted music, stripping out the bits that the record
companies added in. The trouble with this approach is that, once the data has
been extracted, the CIRC encoding is no longer visible. It may not be easy to
tell where the glitches are. For example, it should be possible to create a
low-level but rhythmic distortion that will be noticeable, annoying, and
difficult to identify automatically.
Here's my guess as to why it's harder than just "fixing it in software.":
It sounds like the glitches are being put into the error correction layer protecting the data on the CD. When you read data off of a CD-ROM, the error-correction bits are used to check the integrity of the data, and if a problem is found, it's fixed. All you get is the "corrected data".
While reading audio off of a CD, the decoder uses the bits only as error-detection, and whenever it detects a glitch, it just throws it away - something you wouldn't want to do with data.
In any case, it sounds like by the time you get the data from a CD-ROM drive, those error correction bits are long gone, and detecting that data has been intentionally corrupted is far from trivial.
Now, I may be wrong on all of this - if it is possible to read the "CIRC encoding" from the CD, then it should be fixable in software... But from the sounds of it, this may have to be fixed in hardware.