Sculpture embodies lossy copying using much-copied house-key

Artist Daniel Bejar had a key copied and then a new key copied from it, and so on, until the information embodied in the original key had been lost. He calls the resulting piece "The Visual Topography of a Generation Gap": "A copy was made from my original apartment key, then a copy was made from that copy. This process was repeated until the original keys information was destroyed, resulting in the topography of a generation."

"The Visual Topography of a Generation Gap"(#2, Brooklyn, NY) (Thanks, Fipi Lele!)


  1. A cool looking piece, on an aesthetic and sculptural level.

    In terms of information loss, keys are sort of an interesting choice. It looks like whoever copied the key was doing a (rather rough) analog copy and haemmoraged detail over relatively few generations. Keys, though, are only analog in that they are analog embodiments of their bitting codes, which are not analog.

    If you attempt to reproduce the analog shape, without better machine tools than are to be found in a standard hardware store, you’ll quickly lose it. If you work out the bitting code, you can get the same shape(within the tolerances of the cutting device) as many times as you like. Rather like audio: you can either do lossy copies of an analog version, or suffer the one-time loss of detail by digital sampling and then reproduce that sampling perfectly as many times as you fancy.

  2. If he had correctly adjusted the machine instead of making it copy a 16th of an inch too far he’d have more keys that worked and appeared more intelligent…

    1. yes. It might be intresting to do the same thing with other misadjustments of the key copying machine and see what they look like after 20 generations. But this should serve as a reminder to make a back-up copy of your key BEFORE it is so worn that it isn’t working properly. Of course that won’t help when the pins in the lock become worn.

  3. I am intrigued by the idea of the information loss being exacerbated by the same “error” each time, causing the linear progress seen in the sculpture. The mistakes of the past, being repeated by each new generation.

    I would be curious to see something similar, but with some element of randomness, or data-recovery, or even some sort of adding/changing of “data” over time. I wonder if it would look quite so profound.

      1. It’s worth noting that what you’re linking to is making reference to Alvin Lucier’s sound art piece – ‘I am sitting in a room’.

        As far as I am aware, that’s probably the first example of making art out of the artifacts of lossy reproduction. It’s certainly what occurred to me as soon as I read the headline.

        Also worth mentioning DJ Food’s amazing Raiding the 20th Century mix/audio documentary. If you find this kind of thing interesting, this traces it all the way through to sampling in modern music. It’s been discussed on boingboing before:

  4. “The Visual Topography of a Generation Gap”, or, “Systematic Errors in Machinery: When Humans Fail to Read the Manual”

  5. Lossy copying is a fun project.

    I did a personal project several years ago, where I took a large, complex photo and photocopied it. Then, I rand the copy through the copier. Then ran the copy-of-the-copy through the copier, etc. etc. etc.

    The end result was that the image gradually distorted and began to actually collapse and migrate off the edge of the paper. I ran through an entire ream of paper before I quit. Flipping through the progression (like a flip-book animation) was quite entertaining.

  6. Well, seeing as the key already wouldn’t work after the first 5-10 copies, surely the information has already been lost relatively near the beginning. It’s like a phone number, one missing digit renders it useless, just like the majority of these keys are useless.

    1. Well, yes, that is true. The second key would require some jiggling and by the third it probably wouldn’t work at all. But the point obviously is to show the progression of loss. Plus it looks good.

  7. Fortunately, not all that much information was lost: anyone who has access to this image can now make a very accurate copy of his apartment key.

    1. Biz op: publish some free software that
      takes images of keys and generates
      Gerber (CAD) files. Then you sell generated keys.

      Right now you have to do this with photoshop and/or figuring out your own
      manufacture (eg etch copper-clad and build
      up, stencil and mill, hand-file).

      Seems just right for the grey-hat Makers community.

  8. Confucius say: He who publish photo of apartment key on internet invite many to secret rummage sale. 100% off. Likelihood increase if he user of foursquare.

  9. Yup. There is no way that the artist had someone actually try to copy the keys. It is still a neat sculpture, but definitely a “set up”.

  10. I worked in a key shop as a teenager. In more ways then I’d care to admit it may be the only manual skill I have. I’d bet money these were cut on an automated machine which never work very well. When you get your keys cut, make sure they’re using a manually operated cutter. If they put the keys in the machine and walk away, nine times out of ten your copy won’t work very well.

    Oh, and make sure they don’t put the original where the blank key should go. I made that mistake once. You only need to make it once. Hell is being 17 and having to tell someone that you just destroyed their original key.

    1. You and others are assuming that he didn’t deliberately introduce the errors for artistic effect.

      1. I’m giving him the benefit of the doubt. It’s also entirely possible he made many more copies and only included those that made this pattern. Nothing wrong with that. It’s not a science experiment, it’s art.

  11. I own a key duplication machine. This is a rather forced art project. This sort of error would not occur in the real world.

  12. It illustrates the point well, but the realism is not at all convincing.

    It is the least noisy data set I’ve ever seen, it looks like the “mistakes” were surely intentional and carefully executed.

    At a data level, this looks like a study in linear progression, not information loss.

  13. It does look cool.

    And I think it says that if you put the sort of folks who can’t even copy a key in charge of your culture, you’re going to lose.

  14. 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.

  15. You l know, this is an exceptionally smart idea.

    I wonder if it can be reproduced with audio.

    Each participant would successively tell the next participant what he or she believes to have heard. And at the end of the chain one could publish both statements, the original one, and the one that resulted from it.

    1. That’s called Chinese whispers or telephone. We used to play it all the time as kids.

      Except it never really worked because someone, normally my dad, would pretend to say what they heard, but instead just be making up something that sounded reasonably similar, just because he thought it was funnier.

      Much better is the “Sentence Game” which we still play with friends. Each person writes a sentence, and then passes their paper to the left. When you receive the sentence, you draw a picture best describing the sentence below it. Fold over the original sentence, pass the paper to your left again. When you receive a drawing, write the sentence below and fold over the drawing, pass, and keep repeating. The mutations can become really great.

      1. Somebody has replaced your sarcasm detector with a lossy copy of a sarcasm detector!

  16. I’m guessing he had all the copies made on the same machine. Had he used a different machine each time, the errors would have been more randomly distributed.

  17. This is a very linear progression, which means that the machine produced nearly identical errors on each generation. Thus, the machine adjustment was much worse than the operator’s skill level.

    The lateral displacement error is the most obvious, and indicates that the spacing between the two keyholders doesn’t match the spacing between the cutting wheel and the follower.

    The bumps get lower with each generation, so the height of the follower is also off relative to the cutter.

    There may be some second-order effects in there too, but those are the main errors.

    1. the machine adjustment was much worse than the operator’s skill level.

      Didn’t know there were so many precision machinists and locksmiths reading BB! Get your orange aprons back on and get back to work! The line at the keys desk reaches all the way back to the Lawn & Garden department!

      I’m surprised to see so many people who think this sequence of errors would have had to be deliberately amplified. Maybe your keymaking apparatus is a wonder of aerospace-age precision tolerances, but when I’ve had keys made at the local TrueValue, strip-mall locksmith, or even the Lair of the Orange-Aproned Phantoms, their key-cutting equipment spanned a pretty broad spectrum from soapbar-and-penknife quality to not-so-shabby-for-a-forty-year-old-die-grinder.

      I’m sure the higher-end stuff is much more precise these days, but really, your garden-variety hardware store’s keymaker is intended to make copies from an original that is only a generation or so removed from the Original (or, more accurately, “the last key that worked in that lock”). The tolerances in most household locks have to be at least a little loose, to allow for freedom of movement and at least a little wear and tear over the years. Certainly a key-grinder can be way too far out of adjustment, but I see no reason why this project couldn’t have been made by an acceptably-adjusted machine at any given hardware store. The errors in adjustment would be more-or-less constant in their amplification in a certain direction, assuming the “originals” and blanks were competently installed in each generation. Within 3 generations or so, you’d have a key that no longer worked, and there’s no practical reason to expect an ordinary workaday key-grinder to be adjusted a hell of a lot more accurately than that. By the time you’ve made that many generations of keys (as opposed to a buttload of copies off the same master), it’s time for some new locks anyway.

  18. In theory, I think that this is supposed to mean that, if you make a copy of your key, then make a copy of a copy, eventually you’ll get to a key that won’t work. In practice, I think that you’d get what I usually get when I have a key copy made: the minor errors just causes the key to stick a little when I use it the first few times, but those differences get eroded away by use, so the copy becomes more faithful to the original, or at least faithful enough to use. Eventually, though, I think that you’d have a problem with erosion of the lock, even if it’s made out of much harder metal than the key.

    tl;dr – IANAL[ocksmith]

  19. I’d like to be an artist, but I couldn’t bring myself to dream up wanky phrases like ‘topography of a generation’…

  20. Interesting project, but if it’s supposed to represent the ‘topography of a generation’, it’s missing half the story. The sculpture only represents the topography of what is lost over a generation and assigns no weight at all any new topography that is created over the course of same period of time.

  21. Didn’t Russ Myers make a film called The Topography of a Generation?

  22. Many opinions but not much data. Scientific claims need to be replicated before they can be believed. Has anyone else tried to do this important experiment? I agree with those who are surprised by the apparent monotonic progression in this series – I would have expected something more like a random walk. The smooth curves seem suspicious to me. Even if you assume that the machine introduces some systematic bias whose magnitude varies along the the long axis, it should generate parallel curves, which is not what we see here. Cut the ‘visual topography’ gobbledygook – the burden of proof is on the author to replicate and explain his observation!

    1. I agree with those who are surprised by the apparent monotonic progression in this series – I would have expected something more like a random walk.

      Why would you expect that? Judging from this example, each copy displays a cut that is slightly shallower in depth, and slightly closer to the tip, than the cut in the previous copy. There are other changes as well, which is why the progression is not simply a perfect shallowing of both depth and length of the cut, but those are the two most obvious dimensions of the key cutter’s imprecision.

      If you’ve seen one of these key cutters in action, then you know how they work: the original and the blank are locked into parallel vices, and there’s a cutting wheel which is locked in tandem with a guide tip. The guide tip follows the surface of the original key and the cutting wheel follows those curves and angles, matching the cut. If the guide tip and the cutting wheel are not in precise alignment, the copy will be imprecise. If, for instance, the cutting wheel shrinks a bit (due to the constant abrasion of its cutting surface through usage) and one does not allow for that wear by recalibrating the relationship between the guide and the cutting surface, then the cuts will be shallower in the copy than they are in the original, just as we see in the example. And if the guide’s placement along the key’s lengthwise axis is slightly further down the shaft by 0.01mm than the cutting wheel, then each successive copy will have its groove cut 0.01mm nearer to the tip than the previous copy (assuming, as in the project, that each copy made is subsequently used as the “original” for the next copy).

      In this instance, you’d end up with this very same result. Why would you expect more randomness than that, unless the machine operator was using the machine improperly? I think the point of the whole project is to illustrate that no matter how well-adjusted this particular piece of machinery is, making a series of copies of copies will successively degenerate the image with every iteration until you have nothing.

    2. It’s a smooth progression and no a random walk, because it’s most likely the same machine each time, and so the error it introduces is the same error each time.

      If he used a different machine each time, you’d get something closer to a random walk. However, even then I expect there would be a bias in a specific direction — e.g successive cuts are usually shallower, not deeper — because of intrinsic ways in which the machines can be flawed.

      It’s the same with genetic mutations — some types of errors are much more common than others. A deletion, for instance, is much more likely to occur than an insertion, and this is the case in pretty much every DNA strand on the planet.

  23. It’s annoying that key cutters do that. The number of times I’ve taken keys back because I didn’t realise the new key must be held a little out of the lock to work. Gah!

  24. It’s a lovely piece, but I tend to think in real life that the general trend is that every “copy” is an improvement on the last and that information isn’t lost but added.

  25. I think we should not assume the keys shown are not all of the keys cut in the process. It’s possible the artist only took one out of ten keys cut to compress the concept into one we could easily see.

  26. It’s a METAPHOR, y’all. don’t get so caught up in scientific details. It’s not science, it’s art.

    1. Hey Anon #49 thank you for pointing out that key fact.

      But is this key art key?

      Here’s some on art about a key too, a key song of the early seventies…

      …and the singer almost sings on key, too.

  27. Me likes Anon#18’s comment.
    The artwork here is impossible in real life – no machine would generate these types of cuts.

    Artists are just like people ;-) – when they say something, the would like it to be considered.

    I guess my reply in this conversation to the artist would be:
    Knowledge is lossy. Understanding is forever.

Comments are closed.