Just a nitpick, but...

• Increasing the number of megapixels while keeping everything else the same does not change the amount of light the sensor collects, although each individual pixel gets less light.
• Increasing or decreasing the fill-factor or changing the total sensor size does.
• In low light situations, statistics of the intensity of light should be Poisson, which means that 4 pixels at 1/4 the area, when averaged together, should result in the same amount of SNR, assuming relatively small read noise, which should be dominated by the Poisson shot noise in this situation.
• Thus, the only downside of having just more pixels on a sensor if everything else was equal (note that fill factor of pixels would probably be different between different sensors) is that there's now a lot more bandwidth coming out of the sensor, which could be an issue with power efficiency. This could possibly be mitigated by reducing the bit depth on each individual pixel, if you assume there's going to be more noise.
• One upside is that if the pixel size is smaller than the diffraction spot size, then the relative size of your Bayer mosaic should make demosaicing easier.
• To summarize... if you had more pixels in the same sized sensor and can deal with the extra bandwidth, then you should be able to, in the worst case, downsample to achieve similar or higher performance compared to what you had before.

Don't worry... you have brethren in Japan ;)
http://www.nytimes.com/2009/07/26/magazine/26FOB-2DLove-t.html

If I remember correctly, one of the issues with current three-dimensional displays is that there's a disparity between vergence and accomodation in the eye. That is, normally, you point your eyes inwards to look at something closer, and you focus your lens closer as well. With 3D displays, what happens is that you have to focus at a different distance (i.e. where the actual display is) than what your eyes are verging on (i.e. the apparent depth of the image)... I guess you'll get used to it when you lose accomodation in your eyes as you age... :p

You know... FLAC doesn't actually compress down from WAV all that much. Given current storage sizes, why not simply just sell fully uncompressed audio files? You can use FLAC or whatever as the transmission medium and/or storage server-side to be less of a bandwidth burden, but the user should just see an incoming WAV file, etc. that he can do whatever he wants with...

Personally, I rip all my CDs to --preset-extreme MP3 to listen, since I can just pull out my CD if I really needed a bit-perfect copy (e.g. for voice track extraction).

16-bit audio has around 100dB of dynamic range assuming that the signal peaks. What if you're listening to a source that isn't hitting the peak? For instance, you can have a solo flute or violin part that is playing very softly before the rest of the orchestra kicks in.

Disclaimer: I am not very well versed in statistics and have only just read the Wikipedia article on Instrumental variables.

On page 14, the report discusses the model they use, which is a linear system of the log of the four variables they're trying to find relationships between. They then discuss 4 instrumental variables as well as two dozen or so dummy variables that describe aspects of the actual show, like when it is aired on TV, whether they have related drama CDs, net broadcasts, whether the anime was an original work or based on manga, novels, games, etc., who the target audience was and how many regions the show have been broadcast in. Table 3 has the full list, and also includes what I assume to be slopes in log-log space from their analysis as well as a "t" factor for which I'm not sure about. The four main instrumental (non-dummy) variables are: DVD price (number of thousands of yen per episode, how long a video on YouTube persists ... while the series is airing, within 1 month of that particular episode airing, and after the series is finished airing).

Table 4 is a chart highlighting that there is a negative correlation with DVD price and DVD sales, but a positive correlation of DVD price and YouTube views. There were also positive correlations between how long a video managed to stay on YouTube after airing of the TV show has been completed and DVD sales, rentals and YouTube views, but a negative correlation with Winny downloads.

According to Wikipedia, use of instrumental variables is one way to see if there is actual causation between two variables. However, I'll leave it to someone more well-versed in the subject to see if the report is accurate or not.

Furthermore, you don't need to kill the target for fluorescent microscopy. While attaching labels may affect some of the chemistry in the organism, you can at least still observe biological behavior.

Here's the relevant passage from the article with the juicy bits:

So they took 121 x-ray images of the specimen, with each image taking 2-24 seconds, and then stitched them together using a tomography technique to obtain their 3D volume. It's certainly faster than a few weeks, but this is not what I would consider "immediate". The article also points out that poor cryopreservation led to some artifacts and that the resolution in this technique was still not as good as the TEM; not having an entire 180 degree rotation of the object led to artifacts as well:

So, you're saying she sounds like capsule and perfume? :p

On a more serious note, a (paid) update to Hatsune Miku gives her 6 additional voicing styles. I find that one of them might be able to sound somewhat like a real person if programmed well. Megpoid, using the same Yamaha sound engine, sounds a lot more realistic than the other Vocaloid2 products, and also doesn't require as much tweaking.

It's okay, because she's already under-aged to begin with. (16) :) By the way, Norton Utilities is usually anthropomorphized in Japanese otaku culture as a creepy old man with a stethoscope...

And there's no need for autotune! (Okay, maybe some autotune is in order when Miku refuses to sing the way you're telling her to sing...)

You can think of Hatsune Miku as the artist, or you can think of her as an instrument and the voice manipulator as the actual performer.

The original concert took place on March 9th. This comes out to 3/9, which, if you creatively pronounce it in Japanese, becomes "mi ku". The concert seems to me to be more of a Sega thing than a Crypton event, since the models they used were originally from the production of the PS3 rhythm game. Sitting in the audience was basically the closest you could've gotten to a Sharon Apple experience (sans glitching advanced strike craft).