• (cs)

    A brillant example of the blinding effects of technology. The fact that something is new, unique, and clever means it MUST be better than an old, boring, traditional solution, right? Unfortunately, money speaks louder than common sense. And in this case, the money is saying, "No Quack."

  • Chan (unregistered)

    LoL I cannot stand not to laugh at the section which NN produce

  • Edward Royce (unregistered)

    Hmmmm.

    Come up with a neural network that can do documentation and I'll buy it.

  • (cs)

    What do you get when you put a bunch of idiots together with a REALLY REALLY big roll of duct tape? A duct-tape house (though someone really wanted to tape a poster to the wall)

    Brilliant!

  • Anonymous (unregistered) in reply to Edward Royce
    Hmmmm.

    Come up with a neural network that can do documentation and I'll buy it.

    Sorry, my brain is not for sale. But you may want to ask your friendly neighborhood zombie for a spare one.

  • el guapo (unregistered)

    This neural network has a bright future writing body text for spam.

  • little biologist (unregistered)

    Having both read a book on neural network and having programmed several small programs that parse chess game files (the PGNs) my mind was screaming OUCH OUCH OUCH...

  • Dana (unregistered)

    His really big mistake was trying to reason with management. A more subtle solution would have been to cloak the traditional approach so it would look like a neural network. Judging from management's complete lack of understanding of AI, there's a good chance they wouldn't have known the difference.

    You could coin this one, 'The Lies to Children' technique.

  • ChiefCrazyTalk (unregistered) in reply to Erick

    I call fake. If management was really told there was a better, cheaper way of doing something, and the neural network solution was unworkable, why would by the completely ignore the advice of an actual expert? Never mind, I may have answered my own question.

  • (cs) in reply to ChiefCrazyTalk
    ChiefCrazyTalk:
    I call fake. If management was really told there was a better, cheaper way of doing something, and the neural network solution was unworkable, why would by the completely ignore the advice of an actual expert? Never mind, I may have answered my own question.
    ESPECIALLY if you factor in that a government entity (Bureau of Water Mgt) is involved. That company was suffering stupidity by osmosis. (Pun intended, and sure, there are exceptions to most every rule.)
  • Some Meat-based NN (unregistered) in reply to Dana
    Dana:
    His really big mistake was trying to reason with management. A more subtle solution would have been to cloak the traditional approach so it would look like a neural network. Judging from management's complete lack of understanding of AI, there's a good chance they wouldn't have known the difference.

    You could coin this one, 'The Lies to Children' technique.

    Where's the mistake? Heck, this isn't even a WTF! Here's why:

    1. He identified the customer's requirements and proposed a reasonable solution that would meet those requirements.

    2. The customer identified "must be a solution based on neural networks" as an additional requirement.

    3. He identified the additional cost, in money and time, and explained that this was not a functional requirement.

    4. The customer agreed to the non-functional requirement.

    5. He got paid a shitload of money to develop a system according to the customer's requirements.

    6. He delivered a system that met the customer's requirements.

    7. The next-of-kin customer had different requirements.

    8. He got to implement the system the way he originally envisioned it, according to more sane requirements.

    9. He got paid a shitload of money to do it again, the right way.

    10. He probably learned a LOT about how the real world works; and tricks, tips, and pitfalls about real world implementations of neural networks.

    Sounds like a dream situation to me. Paid shitloads of money to do something I'm interested in; twice. WTF indeed!

  • (cs)
    The pig go. Go is to the fountain. The pig put foot. Grunt. Foot in what? ketchup. The dove fly. Fly is in sky. The dove drop something. The something on the pig. The pig disgusting. The pig rattle. Rattle with dove. The dove angry. The pig leave. The dove produce. Produce is chicken wing. With wing bark. No Quack.
    *clicks his fingers while sipping a cappucino*
  • David (unregistered) in reply to Some Meat-based NN

    Had this been a commercial client, I would have been happy to build them whatever they asked for. Of course most commercial clients count cost as a factor. If I told one of my clients I have an alternative that would save them an order of magnitude (or more) is cost savings by changing their desired approach, they would probably thank me.

    Government entities are a little stranger. The decision makers are not always concerned with cost. Spending available budgets, politics, job security and other concerns often have a greater impact than efficiency and effectiveness.

    I'll bet that the company that hired him promised their client that they would implement this with a neural network. I'll bet that the sales team and upper management were thrilled to get a government contract this big and had to backfill to meet the requirement.

    As a taxpayer, I always cringe when I see government wasting money. When they are wasting the money on me, it's not quite as bad, but still,

  • 604 (unregistered) in reply to Pecos Bill

    I can't help but think he was doing covert work for "the company" and not realizing it. Something like real genius... He was actually training neural networks to starve children in thirdworld countries through gaming the world economic system. He just thought he was doing something mundane.

  • bjhill2112 (unregistered)

    Ah... No Quack. This too is one of my all time favorites, right next to the Paula Bean. Brillant.

  • BJ Upton (unregistered)

    There was a website that had a robot that would do fake Peter Gammons (sportswriter/commentator for ESPN).

    They were actually able to nail his style and create more than a few larps.

  • lol (unregistered)

    no quack. NO QUACK.

  • (cs)

    The real WTF is that "the CTO decided to put the monstrous neural network to use doing something else: prose", becase:

    Neural Networks are not designed to handle tasks like this If developed, such a Neural Network would need a lot of training An algorithmic version could be developed in a fraction of the time and do the job better

    After all, no wonder they ended up with quack.

  • eric76 (unregistered)

    I'd have been tempted to try to find some little itty bitty piece that a neural network might be somewhat appropriate. Do the whole thing conventionally and use a neural network for that one little piece.

    Presto! They would have a neural network solution that works.

  • Dukhat (unregistered) in reply to little biologist

    Ouch indeed. Having done research on neural networks I can't believe such a system was even made to work at all! He must have really known a lot about neural networks to be able to shoehorn them into solving a task they were so unsuited for. I would hire this person in a heartbeat if I had a task that could actually be reasonably solved with a neural network.

  • icanreadneural (unregistered)

    The pig go. Go is to the fountain. The pig went to the fountain. The pig put foot. Grunt. Foot in what? ketchup. The pig tested the fountain water with its foot and found it was in fact ketchup. The dove fly. Fly is in sky. The dove flew in the sky. The dove drop something. The something on the pig. The pig disgusting. The dove shat on the pig. The pig rattle. Rattle with dove. In retalliation, the pig rapes the dove. The dove angry. The pig leave. After doing his business and angering the dove, the pig makes his exit. The dove produce. The dove becomes pregnant to the pig and gives birth. Produce is chicken wing. With wing bark. No Quack. The offspring is part chicken, part dog, no duck.

  • Fr (unregistered) in reply to ChiefCrazyTalk
    ChiefCrazyTalk:
    I call fake. If management was really told there was a better, cheaper way of doing something, and the neural network solution was unworkable, why would by the completely ignore the advice of an actual expert? Never mind, I may have answered my own question.
    That does not really rule it out. This was government contract probably done during the 90s. During that time top goverenment was pushing that awarded contracts include, for big projects, use of some kind of new wow type approach the old computer ways were not good enough. So I am guessing that the contract solution said they would use neural networks as the solution.
  • Cheatah (unregistered) in reply to icanreadneural

    This must be the best interpretation of the poem so far :P

  • German B. (unregistered)

    The story has pointy hair all over it...

  • dribble (unregistered) in reply to icanreadneural

    wow, just....wow

  • (cs)

    Good to see this one back again. Can we have CodeCthulhu once more? Please?

    No quack...

  • Johnny Sutherland (unregistered) in reply to Some Meat-based NN

    Some Meat-based NN - You have summed it up excellently!

  • Pale and Out of shape (unregistered) in reply to Johnny Sutherland

    This is one of my all time favorite WTFs. No quack!

  • Me (unregistered)

    There is no quack! There is no quack! There is no quack! There is no quack!

  • boyter (unregistered)

    Its because of this WTF that I became interested in NN's and used them in my thesis. It really is my favorite.

  • (cs)

    It quacked me up, anyway..

  • (cs)

    the quack is a lie

  • Old name was really worse than failure. (unregistered)

    Interesting to see that the pathetic name, Worse Than Failure was finally abandoned, when the number of readers dropped. That name was really worse than failure.

    This is the first time that I am back since the name fiasco.

  • Guy GUy (unregistered)

    What I dont get is how you get a neural network to produce this kind of output. I can see how you can solve problems such as XOR etc... but this is beyond me. Anyone got a good link for this sort of thing?

  • (cs) in reply to Some Meat-based NN
    2. The customer identified "must be a solution based on neural networks" as an additional requirement.

    IMO part of a system developer's job is to help identify the requirements that might not actually be a requirement, as well as identify requirements they may have missed, rather than just blindly doing as asked.

    I can't see how ignoring the sound advice of an highly skilled expert in the field that you're exploring and charging ahead with a very expensive, highly inefficient system based on one erroneous requirement is not a WTF.

  • bartolo & the nigth breakers (unregistered) in reply to icanreadneural
    icanreadneural:
    The pig go. Go is to the fountain. The pig went to the fountain. The pig put foot. Grunt. Foot in what? ketchup. The pig tested the fountain water with its foot and found it was in fact ketchup. The dove fly. Fly is in sky. The dove flew in the sky. The dove drop something. The something on the pig. The pig disgusting. The dove shat on the pig. The pig rattle. Rattle with dove. In retalliation, the pig rapes the dove. The dove angry. The pig leave. After doing his business and angering the dove, the pig makes his exit. The dove produce. The dove becomes pregnant to the pig and gives birth. Produce is chicken wing. With wing bark. No Quack. The offspring is part chicken, part dog, no duck.

    I feel enlightened. Thanks sir

  • ludus (unregistered) in reply to Guy GUy
    Guy GUy:
    What I dont get is how you get a neural network to produce this kind of output. I can see how you can solve problems such as XOR etc... but this is beyond me. Anyone got a good link for this sort of thing?
    You are putting words next to eachother in ways that are occuring often. NNs can do that if you make them read text for training, and the result is kindof amusing. ;)
  • shd (unregistered)

    Digg it! http://digg.com/programming/Neural_Networks_Surprisingly_Good_at_Writing_Poems

  • S.L.A.N. (unregistered)

    Qui vult decipi, decipiatur.

  • Winslow Theramin (unregistered)

    There's no way this story is real.

  • (cs)

    What was the input that produced that poem as output?

    Bonus points for using the word Zen rather correctly.

Leave a comment on “Classic WTF: No Quack”

Log In or post as a guest

Replying to comment #:

« Return to Article