Comment On No, We Need a Neural Network

M.A. is one of the world's foremost experts on neural networks. His undergraduate specialty was artificial intelligence, his master's thesis was about genetic algorithms, and his doctoral dissertation covered evolutionary programming. Such an extensive computer science education opened up a wide range of career options, ranging from a professor at a university to ... a professor at another university. When someone outside of academia sought out his expertise for a project, he jumped at the opportunity. [expand full text]
« PrevPage 1 | Page 2 | Page 3Next »

Re: No, We Need a Neural Network

2006-12-01 15:15 • by osp70
our tax dollars at work.

Re: No, We Need a Neural Network

2006-12-01 15:16 • by Yuriy
Of course the real WTF will be when the neural network learns too much and goes SkyNet(tm) on it's human masters.

Re: No, We Need a Neural Network

2006-12-01 15:20 • by Anonymous
A wtf with a happy ending? That's a wtf in itself!

Re: No, We Need a Neural Network

2006-12-01 15:21 • by Franz Kafka

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.

 

dear lord, that was almost coherent! 

Re: No, We Need a Neural Network

2006-12-01 15:21 • by bob the dingo
104732 in reply to 104729
i'd like to try these chicken wings the dove produced...and yeah, really, what better use of a server cluster that runs very expensive code than to make a funny for the newsletter!

Re: No, We Need a Neural Network

2006-12-01 15:23 • by Minos
Hundreds of thousands of dollars, and it still doesn't write as well as Mark V Shaney (http://en.wikipedia.org/wiki/Mark_V_Shaney).

Re: No, We Need a Neural Network

2006-12-01 15:23 • by Dan
104734 in reply to 104729

Awesome, This made my day. I'm not going to do what the next X posters will do, which inclues:

(1) Indicating how stupid neural networks are
(2) Indicating that anyone who ever wen to university is a knob
(3) Indicating how much better they could have done it in PHP using Javascript

I'm going to embrace the joy that is pig. And duck. And quack.
 

Re: No, We Need a Neural Network

2006-12-01 15:23 • by J
That prose looks oddly familiar! Why, I think I see that at the bottom of all these emails selling me various drugs!

Re: No, We Need a Neural Network

2006-12-01 15:23 • by Long Timer
104737 in reply to 104730

I admire M.A.'s willingness to help 3 years down the road (I personally wouldn't have saved the proposal for 3 years). He sounds like a stand-up guy.

 I'm still not sure why they trained the system to spew forth prose...

Re: No, We Need a Neural Network

2006-12-01 15:30 • by Older Dude
104739 in reply to 104737
Anonymous:

I admire M.A.'s willingness to help 3 years down the road (I personally wouldn't have saved the proposal for 3 years). He sounds like a stand-up guy.

 I'm still not sure why they trained the system to spew forth prose...

Let's see, M.A. proposed the correct solution, repsected the customer's wishes to do it the wrong way, got the wrong way to work correctly (albeit painfully), politely helped out 3 years later and got another nice gig to do it the right way, and finally did it the right way (again, correctly).

The neural net wasn't the problem here, it was the original management team!

 

I am master of my domain (tm), allowing fools to give me their money for stupidity, and the obvious.

 

Re: No, We Need a Neural Network

2006-12-01 15:30 • by Volmarias
104740 in reply to 104737
Anonymous:

 I'm still not sure why they trained the system to spew forth prose...



Well, the internet is made for two things: Communication, and Porn.

As a government workplace, Porn is not allowed. Therefor, the only thing left is Communication.

Or, to put it more succinctly, "Prose Before Hos"

Re: No, We Need a Neural Network

2006-12-01 15:33 • by bob the dingo
104742 in reply to 104740
Volmarias:
Anonymous:

 I'm still not sure why they trained the system to spew forth prose...



Well, the internet is made for two things: Communication, and Porn.

As a government workplace, Porn is not allowed. Therefor, the only thing left is Communication.

Or, to put it more succinctly, "Prose Before Hos"

 

i do believe this is the reply of the day... 

Re: No, We Need a Neural Network

2006-12-01 15:33 • by biziclop

Sometimes when we face yet another obscure and useless idea from marketing, we dream about building a neural network-driven webapp and then say to them: "We've coded it, now go tell it your ideas."

But now that I know this approach can actually work...
 

Re: No, We Need a Neural Network

2006-12-01 15:34 • by bouk
104744 in reply to 104737

Nice to see an AI related wtf for once. I am sure there must be more of 'em lying around... 

Anonymous:

 I'm still not sure why they trained the system to spew forth prose...

Training a neural network to do something it is not designed to do is cheap and easy, it is the designing that is the hard part.

btw, just for those interested, at the moment text classification is done by support vector machines, which are similar to neural networks, so recognizing what type of input is present in a field might not be so bad after all. But using the neural network to convert the input... *shudders*

Re: No, We Need a Neural Network

2006-12-01 15:34 • by Sergeant

Translation: 

 

They told him to go. Go is to the Water Department. The boss put foot in mouth.

Grunt.

Foot in what? Doo-doo. 

The
MA fly. Fly is in Water Department. The MA drop something. The
something is a design on the boss-pig. The pig is now disgusting. The
pig shake, rattle, roll, and ignore MA.

The MA angry. The boss-pig leave. The MA produce. Produce is BS. With no use.

 

Seriously, no quack.

 

Re: No, We Need a Neural Network

2006-12-01 15:35 • by Older Dude
104746 in reply to 104740
Volmarias:
Anonymous:

 I'm still not sure why they trained the system to spew forth prose...



Well, the internet is made for two things: Communication, and Porn.

As a government workplace, Porn is not allowed. Therefor, the only thing left is Communication.

Or, to put it more succinctly, "Prose Before Hos"

So to best utilize the internet, we need to communicate more porn?

Re: No, We Need a Neural Network

2006-12-01 15:37 • by DrPizza

Alex Papadimoulis:
M.A. is one of the world's foremost experts on neural networks. His undergraduate specialty was artificial intelligence, his master's thesis was about genetic algorithms, and his doctoral dissertation covered evolutionary programming. Such an extensive computer science education opened up a wide range of career options, ranging from a professor at a university to ... a professor at another university.

LOL

Re: No, We Need a Neural Network

2006-12-01 15:48 • by mrprogguy

Congratulations, sir--you've invented a two-year-old!

Allow me to translate: 

In the Near and Not-So-Far-Off Times, O Best Beloved, the Pig went to the fountain, grunted, and put its foot in the ketchup.  (It was, as we know, a ketchup fountain, not a water fountain.)  A Dove flew over and pooped on the Piggie, who was disgusted.  (Ketchup is one thing, poop is another, after all.)  The Pig made an annoyed noise in the direction of the Dove, who became angry.  The Pig departed, and the Dove produced a chicken wing.  The Dove used the chicken wing to make a barking, not a quacking, noise, in the direction of the departing Pig. 

And that is how the Leopard got his spots.

(Anyone know whose style I'm parodying?)

 

Re: No, We Need a Neural Network

2006-12-01 15:48 • by mav
Alex Papadimoulis:


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.

 

 

Oh man, if only I had a nickel for every time I:

Put foot.  Grunt. Foot in what?  ketchup.

 

Its so funny because it speaks to the common man, ya know? 

Re: No, We Need a Neural Network

2006-12-01 15:53 • by Aaron Griffin
brillant!

Re: No, We Need a Neural Network

2006-12-01 15:57 • by The dove angry, the pig leave
Hah, that was great. The nonsense poetry at the end made it _that_ much greater.

I'm amazed that they actually managed to put together a NN that did the job satisfactorily once trained, the task sounds like it fits a NN approach extremely badly.

Re: No, We Need a Neural Network

2006-12-01 15:58 • by Anonymous
Fellows, I think we've got something here that will pass the Turing test. Test in what? Ketchup.

Re: No, We Need a Neural Network

2006-12-01 16:03 • by Theologian

The dove drop something.

Ewwwwwwww.......

Re: No, We Need a Neural Network

2006-12-01 16:04 • by obediah

I've never felt jealous about a WTF before.

 

Re: No, We Need a Neural Network

2006-12-01 16:17 • by biziclop
104757 in reply to 104753

Anonymous:
Fellows, I think we've got something here that will pass the Turing test. Test in what? Ketchup.

Have you ever wandered what would a system be like if it was designed to pass the Turing test? Paranoid, obsessive and monomaniac. Scary. No quack.

Re: No, We Need a Neural Network

2006-12-01 16:24 • by EvanED

Alex Papadimoulis:
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.

 Haha! That's great!

But actually, I wonder what the source data was for this. Bceause it actually seems fairly impressive to me... I'm curious how this actually worked.
 

Re: No, We Need a Neural Network

2006-12-01 16:29 • by ParkinT

The machine finally did what it was designed to do!

 This story reminds me of Business Spew [The BS generator]  A clever application written by a dear friend (brilliant guy who succumbed to Leukemia last year and died at 40 something).  Working in "Corporate America", Wilson would tell stories of how he posted pages of nonsense produced in this way on the company bulletin board and watch people huddle around it, discussing the "true meaning"...

I had to dig in The Wayback Machine to find his site.

Re: No, We Need a Neural Network

2006-12-01 16:32 • by ParkinT
Not enough XML !

Re: No, We Need a Neural Network

2006-12-01 16:34 • by elbaz

This is a classic management by trade magazine scenario.  An old boss of mine used to send the airline magazines down with articles highlighted; it was up to me to explain the idea and how it would work, or more often then not, not work with our system.

 

On another note; you have to wonder if the previous management team was financially related to the hardware vendor.

 

Re: No, We Need a Neural Network

2006-12-01 16:35 • by jggles
It might have actually been quicker to write a GA to evolve the net to do it.

Re: No, We Need a Neural Network

2006-12-01 16:36 • by wk2x
104763 in reply to 104758

I think "no quack" may very well become the next "brillant."


Maybe Alex should see about putting it on a mug ...

 

 

Re: No, We Need a Neural Network

2006-12-01 16:36 • by gl
104764 in reply to 104749
mrprogguy:

(Anyone know whose style I'm parodying?)

 

Ummmm.... What-The-Faulkner?

Re: No, We Need a Neural Network

2006-12-01 16:41 • by GrandmasterB

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.

 Jeesh, the thing writes more coherently than many college freshmen.

capcha: clueless (as-if!) 

 

 

Re: No, We Need a Neural Network

2006-12-01 16:45 • by Charles Shapiro
104768 in reply to 104733

Wow, I hadn't heard of Mark V Shaney. Pretty cool.

 

I
was thinking more of racter ( http://en.wikipedia.org/wiki/Racter) I've
got an original copy of _The_Policeman's_Beard_is_Half_Constructed_.
It's a prized possession.

Re: No, We Need a Neural Network

2006-12-01 16:45 • by foonly
104769 in reply to 104744
bouk:

Nice to see an AI related wtf for once. I am sure there must be more of 'em lying around... 

 

Yeah.  Back in the day I worked on an automatic dispatching system.  My boss decided not to reinvent the wheel, so he payed an enormous (for then) amount of money to buy the source code for the automated dispatch system designed for [A CITY].   After weeks of waiting we get in ... 20K lines of

if (200 yards past x) do y

else if (...

Not one line of general purpose code in the whole system.  The contractor had just sat next to dispatchers and written down exactly what they did, then they coded it.  ANy change to the system would have been very expensive.  Good thing it was never used.

Re: No, We Need a Neural Network

2006-12-01 16:52 • by foxyshadis
I like. Thread new memes in ketchup. Memes rattle. No quack.

Re: No, We Need a Neural Network

2006-12-01 17:03 • by Jake

Pretty damn funny, I likey this one

 

captcha = whiskey

Re: No, We Need a Neural Network

2006-12-01 17:20 • by mbvlist

In the Netherlands, there is this rather 'odd' == WTF descision of the highest court, about a similar case. Some advisor said: "Do A, don't do B because it will go wrong at this list of points". Customer: "But we want B. Please help us to prevent that list of point from occuring while implementing B". The advisor agreed.

Years later, when things have gotten really messed up, they sued him for doing B. Proof was not the issue: all decisions and such were well documented. He did B, and B went wrong, so sue this guy. And they won 

That is the biggest WTF for this court I know of, and something all advisors in the Netherlands should know. Unfortunately, I don't think they do...
 

Re: No, We Need a Neural Network

2006-12-01 17:26 • by jimlangrunner
104774 in reply to 104768

Management WTF.

We need more new tech

No Quack!

 

Re: No, We Need a Neural Network

2006-12-01 17:30 • by mrprogguy
104775 in reply to 104764
Anonymous:
mrprogguy:

(Anyone know whose style I'm parodying?)

 

Ummmm.... What-The-Faulkner?

 Good guess, but no quack for you.  It's Kipling.

Proper answer to the question "Do you like Kipling?" :  "I don't know, I've never Kippled."

Re: No, We Need a Neural Network

2006-12-01 17:33 • by Brad

Please, please, please, please can I have more prose?  Seriously, I would love to get that newsletter.  Even just an bi-weekly email.
 

Re: No, We Need a Neural Network

2006-12-01 17:47 • by Vector
Well, I guess the neural network had it's prose and cons.

Re: No, We Need a Neural Network

2006-12-01 17:48 • by Onanymous
104778 in reply to 104776
Anonymous:

Please, please, please, please can I have more prose?  Seriously, I would love to get that newsletter.  Even just an bi-weekly email.
 

Seconded!

Re: No, We Need a Neural Network

2006-12-01 17:54 • by Rob H

For text segmentation, something like a CRF (Conditional Random Field) or HMM (Hidden Markov Model) is more appropriate than an ANN, but if the text is in a well known format such as CSV (as opposed to pure natural language) then obviously these things are not necessary.

Its a shame that the managers didnt want to use the machine learning techniques more appropriately (e.g., for dimensionality reduction, pattern recognition etc.)

 

Re: No, We Need a Neural Network

2006-12-01 17:58 • by ha ha
It took three years to fire them?

Re: No, We Need a Neural Network

2006-12-01 18:32 • by Akaji

I'm a second year CS major at a private university in Minnesota, and our professor had an absolutely great idea for a project - a random poetry generator.  Instead of randomly stringing together words, however, the generator could only put words together that appeared together in the 'original' text (a poem written by a famous author was put into a .txt file, and the program would analyze it).  I've gotten some pretty similar results to what that NN got...

 For example, using a Lowell poem as the input, I got the following (this is just two short snippets of two run-throughs of the program):

The stone statues
of the antiquated refrigerator gurgled
mustard gas through your meals
tasting like a ball of
the blood-flood: white churches hold
your first kill; its hackneyed
speech, as a tin wastebasket
of the chalk-dry and vegetating
kingdom of sparse, back.. ."

red; my ill-spirit sob in a ball of the paranoid,
you hold their underworld garage.. Space is nearer. And there,
the morning with orange, she buys up tons of Coleridge,
hull to burst the garbage pail, she buys up tons
of morning, propped its hackneyed speech, blossoms on the Rahvs
in a black. Parking spaces luxuriate like a ball of
Coleridge, you turn your throat.. The bronze weathervane cod has
had rushed it cooled our long maneuvered visit from the
Masters of the fish. You said:

  

 Good times. 

Re: No, We Need a Neural Network

2006-12-01 18:34 • by Kippler
104783 in reply to 104775
mrprogguy:
Anonymous:
mrprogguy:

(Anyone know whose style I'm parodying?)

 

Ummmm.... What-The-Faulkner?

 Good guess, but no quack for you.  It's Kipling.

Proper answer to the question "Do you like Kipling?" :  "I don't know, I've never Kippled."

You should have let someone else have a try, I totally knew that one. >:(     

Re: No, We Need a Neural Network

2006-12-01 18:35 • by Paul Harrison

Correct response to such a request is:

 

Re: No, We Need a Neural Network

2006-12-01 19:01 • by Cookie
104786 in reply to 104744

support vector machines, which are similar to neural networks

Similar ?

Oh, you mean perhaps that they are both about AI... 

Re: No, We Need a Neural Network

2006-12-01 19:02 • by dave's not here
Something is NQR about this scenario. It's very easy to paint this as a management WTF, but more likely someone in management wanted this guy or project for some other agenda. It's possible they expected to apply this project to some other area outside their immediate responsibility, but had to use this scenario in order to justify the cost. I can just imagine some manager thinking they can piggyback their career on some 'breakthrough' technology.

Either that or M.A. is a really bad communicator.


« PrevPage 1 | Page 2 | Page 3Next »

Add Comment