Crap you beat me to it... Oh well, as I like to think, if an idea that occurs to me has already occurred to someone else, it probably wasn't such a great idea anyways!
The problem is that Cuil is starting off by claiming that it's 2,784 better than it really is. That's as if your 17 year old claimed that he's done more and better than your 27 year old, which is what Cuil is claiming.
Of course you realise that Google actually has these 4,320 pages, whereas Cuil just lies by blowing up their number out of thin air by 2,784. You surely also realise the difference between going "We have 4,320 pages for your search, but well only 7.4% of them are non-redundant" and "We have 2,784 pages for your search, but only 0.036% of them are (well, is) non-imaginary".
Man, what are you talking about. We're talking about STFT/MDCT based compressions, which is what YouTube uses. That means it stores frequency-domain bins, not samples.
As for what you talked about anyways, 16 bits sampling that only uses the upper 8 bits? What kind of compression formula would get you to that?
huh?? And why would it do that? You do realise that the compression talked about here makes pretty much everything loud, right? Therefore if you omit something it'll be even more noticeable.
2784 is an integer multiple of 3, so it's more like 1 in 3, but then it's unwise to generalize from a data set where n=1.
Bzzzz wrong. For the result of a division by 2784 to be an integer, the numerator has obviously to be a multiple of 2784. Therefore I'm right in saying 1 in 2784. I accept your apology.
lol, right, a bug. More like they want to sound like a contender, and have an edge over the king. However if I'm correct and they only have 43 million pages, that's less than Google in late 1998 (60 millions). No wonder why they'd want to inflate their numbers.
As for being an honest mistake, no way, it's easy to know whether your database contains 120 billion pages or 43 millions.
Why does Cuil claim to have "2,784 results" to my search yet display only one? Does it mean we have to divide the impressive 121,617,892,992 claimed index web pages by 2,784 to obtain the astoundingly round number of 43,684,588? What are the odds that the result of this division would be an integer number?
The best they can hope for anyways would be to be bought by Google. Either that or they'll stagnate due to scalability issues or even suffer a slow death.
Besides, "Search 121,617,892,992 web pages" and none from my website? Allow me to remain sceptical..
Oh that's not the same clipping we're talking about. The clipping we're actually talking about is when sample values get hard-limited, which creates distortions (also referred to as non-linearities because they emerge from having the sound go through a non-linear system, in this case a function that forbids a sample value from exceeding a certain value) and can make even ultra-sounds produce audible distortions, which is one of the two main principles behind the so-called "sound laser".
Wow, I didn't know American Rednecks actually had the skill to even use a computer let alone get a Slashdot account;).
Firstly it's kind of racist (yes, Rednecks originally refer to Irish and Irish-Scot immigrants) and secondly he didn't get a Slashdot account, he was an anonymous coward.
Normalising has no effect on the sound, other than an overall gain change.
It's not what I'm talking about. I'm talking about "volume normalization" as VLC calls it, which matches to the description I gave of it.
Also the "compression" you're talking about is the "volume normalization" I was talking about. I thought that compression was simply accomplished by directly applying a function to the sample values?
Slashdot Mirror
According to this thread that's grammatically way not wrong enough to deserve an intervention from the grammar Gestapo.
4 Watts? Wait.. doesn't a ARM Cortex-A8 only use 300 mW?
Is it like square rooting samples?
Interesting, however I was rather thinking of putting it like straight against the very hot part of the reactor.
Crap you beat me to it... Oh well, as I like to think, if an idea that occurs to me has already occurred to someone else, it probably wasn't such a great idea anyways!
The problem is that Cuil is starting off by claiming that it's 2,784 better than it really is. That's as if your 17 year old claimed that he's done more and better than your 27 year old, which is what Cuil is claiming.
Well I sent a mail to these guys, showing them this discussion and asking for their comments, so we'll see ;).
Of course you realise that Google actually has these 4,320 pages, whereas Cuil just lies by blowing up their number out of thin air by 2,784. You surely also realise the difference between going "We have 4,320 pages for your search, but well only 7.4% of them are non-redundant" and "We have 2,784 pages for your search, but only 0.036% of them are (well, is) non-imaginary".
Man, what are you talking about. We're talking about STFT/MDCT based compressions, which is what YouTube uses. That means it stores frequency-domain bins, not samples.
As for what you talked about anyways, 16 bits sampling that only uses the upper 8 bits? What kind of compression formula would get you to that?
huh?? And why would it do that? You do realise that the compression talked about here makes pretty much everything loud, right? Therefore if you omit something it'll be even more noticeable.
What innovation? The iPod?
The iPod, the iPhone, the iMac, Mac OS X, every Macintosh computer since 1997, the Macintosh, the Apple II.
Yeah really, what has that slacker done for Apple anyways? If he hadn't returned to Apple I'm sure Apple would be thriving now. Or dead.
2784 is an integer multiple of 3, so it's more like 1 in 3, but then it's unwise to generalize from a data set where n=1.
Bzzzz wrong. For the result of a division by 2784 to be an integer, the numerator has obviously to be a multiple of 2784. Therefore I'm right in saying 1 in 2784. I accept your apology.
lol, right, a bug. More like they want to sound like a contender, and have an edge over the king. However if I'm correct and they only have 43 million pages, that's less than Google in late 1998 (60 millions). No wonder why they'd want to inflate their numbers.
As for being an honest mistake, no way, it's easy to know whether your database contains 120 billion pages or 43 millions.
This
Why does Cuil claim to have "2,784 results" to my search yet display only one? Does it mean we have to divide the impressive 121,617,892,992 claimed index web pages by 2,784 to obtain the astoundingly round number of 43,684,588? What are the odds that the result of this division would be an integer number?
1 out of 2,784.
The best they can hope for anyways would be to be bought by Google. Either that or they'll stagnate due to scalability issues or even suffer a slow death.
Besides, "Search 121,617,892,992 web pages" and none from my website? Allow me to remain sceptical..
Oh that's not the same clipping we're talking about. The clipping we're actually talking about is when sample values get hard-limited, which creates distortions (also referred to as non-linearities because they emerge from having the sound go through a non-linear system, in this case a function that forbids a sample value from exceeding a certain value) and can make even ultra-sounds produce audible distortions, which is one of the two main principles behind the so-called "sound laser".
Wow, I didn't know American Rednecks actually had the skill to even use a computer let alone get a Slashdot account ;).
Firstly it's kind of racist (yes, Rednecks originally refer to Irish and Irish-Scot immigrants) and secondly he didn't get a Slashdot account, he was an anonymous coward.
Elitist.
Did you even read TFS? We're talking about a 19 kHz sine louder than the rest of the sound.
Well of course he can bend spoons with his mind!
OK, but I was talking about the very problem at hand here.
OMG! Can you say killer app?
Could it be used to get more power out of a nuclear power plant?
Nah unfortunately it does it on the fly, and thus comes with an annoying delay.
Normalising has no effect on the sound, other than an overall gain change.
It's not what I'm talking about. I'm talking about "volume normalization" as VLC calls it, which matches to the description I gave of it.
Also the "compression" you're talking about is the "volume normalization" I was talking about. I thought that compression was simply accomplished by directly applying a function to the sample values?