Lines of code are the wrong approach
Webneurons Web 3
  Lines of code are the wrong approach. An email to Jason Larnier    

Webneurons Web 3 Home
Back
Page titles A-Z index
Main Pages
 Original Posting
 Original Site
 Newsgroup Post
 Web 3
 Wrong Approach
 XML
Reference
 Links
 Contact us


Webneurons Web 3 - The main goal of webneurons is to develop a better way of programming computers. Demonstration of a simple webneuron network. Also includes a minimal javascript XML parser of 37 lines.


Lines of code are the wrong approach. An email to Jason Larnier

Date: Sun, 22 Oct 2006 23:37:42 -0400
From: John Middlemas
To: Jaron Lanier
Subject: Lines of code = wrong approach

Dear Jaron Lanier,

I read your comments at http://sss.stanford.edu/others/jaronlanier/ that current programming methods are not up to the AI task.

"Book programming" is the big mistake. Most all languages use a book like system with lines and pages of code. This is a natural assumption of academics who are used to reading books but after 50 years we still haven't got strong AI so the basic assumptions ought to be questioned. The first thing you need in anything is the right tools.

Most languages use a collection of sub-units whether it be subroutines, functions, objects, classes etc. Sub-units seems to be the right approach and fit with how the brain works where the sub-units are neurons. A big task needs breaking down into sub-tasks. What is questionable is the way the sub-units talk to each other. With line-of-code programming you can't easily find you way around in a big system even with OOP.

The sub-units should be hyperlinked just like the web page success story. Replace the page based lines-of-code system with a web browsable map of hyperlinked network sub-units. You may still want some lines of code within each unit but the connection and data communication between units should be by hyperlinks. If the units were small enough even basic jargon like if() and while() could be replaced by linked collections of sub-units. As for ease of editing, it can all be done by point and click rather than scouring down lines.

However, the main advantage of a network programming system is not just the elimination of jargon and the lines-of-code method but the addition of new features such as ability to create reusable complete collections of linked sub-units on the fly (sub-structures), rather like generating a mini-website on the fly using php, except the sub-structure can have inputs and outputs to other sub-units/structures. Large structures like multi-dimensional spreadsheets could be made on the fly or extended. You wouldn't use MySql, the data would be in the sub-units and indexed however you like.

So all the common applications like database, spreadsheet, and ultimately AI would come under the same basic network programming system. The sky is the limit because the network method easily opens up all the possibilities you could imagine rather than being constrained by structured programming methods. It is freedom.

I guess in some ways the old GOTO programmers had their finger on it but couldn't finish the job.

I am currently working on this and if you are interested further please let me know. It is a big task and I don't know if I can complete it.

Sincerely,
John Middlemas