wu :: forums - Print Page


    
      
        wu :: forums
        (http://www.ocf.berkeley.edu/~wwu/cgi-bin/yabb/YaBB.cgi)
      

        riddles >> general problem-solving / chatting / whatever >> foolishness of "natural language programming&
        
(Message started by: nakli on Nov 17^th, 2010, 4:14am)

Title: foolishness of "natural language programming&
Post by nakli on Nov 17^th, 2010, 4:14am

http://www.cs.utexas.edu/users/EWD/transcriptions/EWD06xx/EWD667.html

Now Edsger W.Dijkstra is a wise man.
His such discussion about other things like algol, ada etc. have proven to be true.

I respect the man but disagree here. I have no solid reason, just that I wouldn't like to believe something as impossible.

He makes some good arguments. Thought we might discuss.

Title: Re: foolishness of "natural language programm
Post by towr on Nov 17^th, 2010, 2:30pm

He doesn't say natural language programming might not be possible. He mostly says that it may not simplify the programming as much as people think it would. And I think he's right; because most people have trouble articulating what they want even in natural language. So you would have a computer taking a guess at what they might want, and since it isn't clearly specified, it will go wrong pretty often. (Which isn't to say it won't ever go wrong if you have a programmer in the middle translating a client's need into software, but at least the computer will do what it is told and not take a guess at it).
I also think he's right that formalisms actually make it easier and faster to tell a computer what exactly you want it to do. Between two people that understand the language of mathematics, a formula is often a much more concise and clearer way to share knowledge than trying to put it into just words.

Title: Re: foolishness of "natural language programm
Post by JiNbOtAk on Nov 17^th, 2010, 10:33pm

A very interesting read. I had to agree with the author, using natural language for programming might be possible, but it would very difficult to get consistent results. A simple example can be seen in examples of formal letter writing; different authors have different opinions on the do's and don'ts. And that is just writing letters, imagine having a programming language.

Even with AI, I don't think it's going to be that easy. Different people speak differently, there are nuances, similes that come into play. So, unless there's a way to actually pick someone's brain and understand what he or she really means..we're pretty stuck on the available languages.

Quote:

..some people found error messages they couldn't ignore more annoying than wrong results,..

So true, back when we dabbled in programming, error messages are usually punctuated with colourful words from the programmers. Our tutor thought it was funny, there were some words we invented just for a particular error message.

Title: Re: foolishness of "natural language programm
Post by nakli on Nov 18^th, 2010, 2:38pm

Author died in 2002.
Any idea when this might have been written?
( Some say around 1978 )
Is it important?

Title: Re: foolishness of "natural language programm
Post by towr on Nov 19^th, 2010, 3:31am

Considering the numbering, and the archive at http://www.cs.utexas.edu/users/EWD/index06xx.html I'd say 1978 is probably right.
I don't think it matters much with regards to the arguments, since they don't depend on the state of technology. Formalisms also help to ease human-to-human communication, by adding precision and clarity. So even if a computer had the level of intelligence needed to understand what its human-language-programmer was saying, it would still hold true that using a formalism would probably be both more precise and faster. Natural language is ambiguous, and to clear that ambiguity up you need to use more language.

Title: Re: foolishness of "natural language programm
Post by ThudanBlunder on Nov 20^th, 2010, 6:16am

Interestingly, in the previous paper (http://www.cs.utexas.edu/users/EWD/ewd06xx/EWD666.PDF) he relates how he solved the product/sum metapuzzle in his head. :o

Title: Re: foolishness of "natural language programm
Post by nakli on Nov 21^st, 2010, 8:11am

on 11/20/10 at 06:16:46, ThudanBlunder wrote:

Interestingly, in the previous paper (http://www.cs.utexas.edu/users/EWD/ewd06xx/EWD666.PDF) he relates how he solved the product/sum metapuzzle in his head. :o

...and the sister in the sidenote !

Title: Re: foolishness of "natural language programm
Post by Grimbal on Dec 1^st, 2010, 6:53am

In that paper he says:
"Later I heard that some people when confronted with it, instead of solving it themselves, with or without pencil and paper, went to such extremes as writing a computer program for it!"

Extreme? Well... I thought that is the most natural thing to do.

Title: Re: foolishness of "natural language programm
Post by jonydec on Jun 17^th, 2012, 9:39pm

Whether they're better or not is viewpoint, but that looks like some mutated combination of COBOL and BASIC, which is most definitely epically bad.

So in my view, no. I think somewhat to-the-point dialects that still use legible verbs/adjectives/names are better (C++, C#, PHP, etc are my recommended languages).

Some dialects begin to get too high-level and/or verbose, creating the real sense so abstracted it's difficult to know what does what. Some are too low-level and brief, pushing you to clearly condition everything you want done. A stability between legibility and brevity, with energy and versatility, is what is best for growth.