Extraordinary claims. Ordinary investigations.

The Evolution of Noise

“Open the last saved jpeg image. Save it as a new jpeg image with slightly more compression. Repeat 600 times”. The result, illustrated above in Generation Loss by David Elliott, is clear: the final image is almost unrecognizable, as something akin to noise takes over the whole frame. A sort of “digital noise”.

The JPEG image format, ubiquitous from webpages to digital cameras, is a “lossy” compression format. Information is irreversibly discarded, lost in the process, in return for a huge space saving and a nearly imperceptible apparent degradation. But as Yoda would say, degraded it is, as information has been lost and digital artifacts have also been introduced.

The simple lesson is that you should never re-save an image in a lossy format like JPEG, unless you are willing to lose information in the process. If you received the image already as a JPEG, recompressing it would not even save that much space, as compressing an already compressed JPEG will not do much. It will not degrade the image as extremely as shown in the video above, since most of the degradation there was due to increasing the compression – without it, recompressing a JPEG even 2000 times will not turn it into complete rubbish – but degraded it will be.

As ubiquitous as JPEGs are the MP3s, this gift from the computation gods that revolutionized the entertainment industry, much to its disgust. This compression format is also lossy. It discards information by the buckets, and you may guess where we go from here.

What happens if you open the last saved MP3 and save it as a new MP3, 600 times? Something like this:

The music is unmistakable, but shrieking noises are a torture. Worse (or better) than that is recompressing Iron Maiden’s “The Number of the Beast”… 666 times, in something like digital Satanism. Click below to listen to the results:


Can you hear the whispers from hell? Probably not, but all these noises and experiences can mean much more than you imagine.



The metallic noises from the Satanic MP3s may have sounded familiar. In bad quality cell phone calls and Skype conversations you may have heard them. Though they don’t use MP3 compression, they do pass through lossy digital compression, with effects, or artifacts, not that much different.

These digital artifacts resulting from a lossy compression begin to accumulate over the successive generations until they overcome the original signal. As they are digital and the algorithms are deterministic, the losses and artifacts are not random, and therefore not exactly “noise”, but if you saw and heard them, they are a sort of “digital noise”. We will get back to this peculiarity later.

Real noise is surely very easy to come up with analog systems, however. Patrick Andrews photocopied an image of himself 100 times, and got this:


If  you paid attention to the little details of any photocopy you may have found these blobs familiar. The toner shows some “artifacts” by aggregating into those globs, but those are usually not very noticeable. The recursive copies made them bigger. Very big.

And in 1970, composer Alvin Lucier performed an amazing work of art. Titled “I Am Sitting In A Room”, it consists merely of his voice dictating the following self-explaining text:

"I am sitting in a room different from the one you are in now. I am recording the sound of my speaking voice and I am going to play it back into the room again and again until the resonant frequencies of the room reinforce themselves so that any semblance of my speech, with perhaps the exception of rhythm, is destroyed. What you will hear, then, are the natural resonant frequencies of the room articulated by speech. I regard this activity not so much as a demonstration of a physical fact, but more as a way to smooth out any irregularities my speech might have."

The result:

Creepy, isn’t it? And also fascinating. I’m not sure if what we hear in the end are indeed the resonant frequencies of the room and not simple microphony.

Either way, Lucier’s work became a cult classic. And Evan Borman made this video version:

Watch as the video, reproduced in a TV screen and then recorded repeatedly with a video camera, quickly degrades into the flickering between the image formed in the screen and the frame capture rate of the camera.

All too complex? But there is something very simple and extremely important that all these examples of recursive copies demonstrate. All of them degenerate into the noises and particular artifacts, initially almost imperceptible, of the systems used to copy them. The original is almost completely lost. The medium (noise) becomes the message.

Small JPEG artifacts initially seen only around magnified high contrast areas spread across the image and formed a series of huge Mondrianic abstract patterns. Metallic blips and shrieks in MP3s become a a long excruciating cacophony akin to whispers from hell. Small blobs of toner become huge blobs. Echoes and microphony took over a dictated voice, as flickering bars dominated a TV screen.

The noise took over the copies. Why is this important?



Recursion is the process by which a process is applied over itself, repeatedly. Recursive copy is therefore the copy of a copy (of a copy, of a copy…). Instead of copy, however, I could have used the word…

Reproduction. Recursive reproduction. And reproduction may remind you of something.

We reproduce ourselves. Animals reproduce themselves. Plants reproduce themselves. Living beings usually reproduce themselves. Prolifically. And the reproductions also reproduce themselves. Life involves a long series of recursive reproductions.

Biological reproduction seldom is, if it is at all, “perfect”. Not only because the child of Albert Einstein and Marilyn Monroe could have ended up with his looks and her brains, but because such an offspring would not be the simple recombination of the genes from both parents. Especially in complex organisms such as ours, there are always some errors in the reproduction, also known as mutations. Most of them are innocuous, but as Yoda would say, occur they do.

There are some noises introduced in each reproduction, and by now you know, after so many examples, what happens after so many successive recursive reproductions. The “noise” take over the original signal.

That’s a very simple and illustrative analogy for the aging process. As each cell in our body reproduces, even if they do not involve any recombination and are supposed to be exact copies, the reproduction is never perfect and the errors accumulate with time. At some point the reproduction will be so far off that it will not be recognizable. Usually the reproduction stops working far before that. We age. We eventually die. It’s an inevitable consequence of life.

But beyond the analogy with the aging process, there’s another important mechanism that involves recursive reproductions and radical changes over many generations. It’s known as… evolution. The evolution of species that we celebrate in this Darwin Year.

Of course, evolution is not a process of degradation. Contrary to what many assume, it is neither, necessarily, a process of enhancement, of progress. Evolution is in the broadest sense simply change. Not necessarily change for good, nor bad. Simply change.

And recursive reproduction surely leads to change. As we have seen, it’s inevitable. Evolution is a fact, and photocopying a photo 100 times is a very didactic demonstration of this inevitability. Unless Creationists claim that somehow all living organisms have developed some divine (haha!) way of perfect reproduction free from any transcription errors, then evolution is simply inevitable through the countless recursive reproductions that perpetuate life itself.

Granted, how evolution leads from unicellular organisms to complex ones like us is something that this analogy doesn’t help that much in clarifying, as there’s no selection of the changes that occur during the successive copies that we presented here.

Or maybe there is, in a way. The errors in the copies was not random, especially not so in the digital ones. In the case of the JPEG image with progressive compression levels, artifacts became noticeable and then started to multiply over themselves, until everything was artifacts. The algorithm that produces and, may we say, favors these artifacts is deterministic. And though the artifacts added are just slightly noticeable, recursive reproduction shows the power that it has in making small cumulative selections very, very noticeable.

Alright, that may not have helped that much, and this analogy is far from perfect, but it does suggest a point of view by which we are almost pure “noise” produced by zillions of recursive reproductions of a single original organism that emerged billions of years ago. It is provocative.

That would be a very special sort of “noise”, one that is not necessarily worse, or even better, than the original. Just different. And a noise produced and selected through all these generations in countless different, and not at all random, ways. It would not present evolution as a linear progression of individuals, but as the recursive reproductions of a singe original unicellular organism, reproductions that changed so much…

The original single-celled organism is still speaking “I Am Sitting in a Room”, and after billions of years, the result is us. Provocative it is.

– – – [via BoingBoing, Reddit, Microsiervos and with thanks to girino]

Blog Widget by LinkWithin

Popularity: 1% [?]

Posted in Science | 6 comments

6 Comments so far

  1. The Evolution of Noise April 1st, 2009 7:29 am

    […] is the original post: The Evolution of Noise Contributor: […]

  2. rivets (Lindsay Marshall) April 1st, 2009 8:50 am
  3. The Evolution of Noise | forgetomori April 1st, 2009 2:45 pm

    […] via The Evolution of Noise | forgetomori. […]

  4. […] The Evolution of Noise | forgetomori. […]

  5. neverville (A Johnson) April 3rd, 2009 4:56 pm
  6. GSGill August 16th, 2009 10:27 pm

    Hmm, in the case of evolution, information isn’t actually lost, it is merely changed. Lossy Image compression, “I am sitting in a room”, they are both examples of lost information. There are no mistakes involved – if you compress the same image 600 times on 600 different computers, you will get the exact same result, UNLESS some mistake occurred during compression.
    I think a much better analogy of evolution ( on a genetic basis ), would be copying the same image, using an unreliable method, many times over. There will no doubt be mistakes introduced in every copy. Hence the end result may be somewhat different than the original.
    I guess both are examples of noise, but I don’t think evolution is a “special” kind of noise.

Leave a reply

Live Comment Preview