In the age of AI, is writing clean code worth it? Essay by @CodingComputing

In the age of AI, is writing clean code worth it?
=================================================

Author:@CodingComputing			Word Count: 923

---

Artificial Intelligence (AI) is transforming all walks of life today,
and the field of software development is also feeling this wave of
change; AI is now able to assist in writing code. A valued skill in
software development is to write clean code: code that is easily
understandable and maintainable. Such developments in AI lead to a
question, that, if AI assists in writing code, does it make clean code
obsolete? In this essay, I explore the ways in which AI can help in
coding, and argue that writing clean code is very much a necessity, in
spite of AI assisting us in coding.


Let us take a closer look at some characteristics of AI, and of clean
code.

ChatGPT, a recently developed AI system, has demonstrated capabilities
of producing code, commenting on code, and even explaining the working
of certain code snippets. Although it is capable of remarkable feats, AI
is limited in certain respects. The limitations are:
1. it does not have real-life experiences,
2. it does not have a fundamental understanding of the world.  Due to
   these limitations, it can only process its training data, and
   synthesize it as per the input request.

Clean coding is about understanding and following certain principles,
that make the code easier to read and maintain. The principles include
practices such as factoring code appropriately, and using correct
indentation. Writing clean code takes time, effort, and discipline.
Developers often shy away writing clean code, and ultimately pay the
price of dealing with bad code [1].

The question we are trying to address is: Now that AI can assist in
writing code, do human developers still need to learn and follow the
principles of clean code?

Let us examine in what way AI can assist in writing code. Primarily, AI
can help with the following tasks, which might raise the question of
relevance of clean code:
1. Producing code from a given prompt
2. Explaining what certain pieces of code do

Let us now explore some objections against clean code that can arise
from the above points, and address the objections.


1) AI can produce code for us
-----------------------------

Some might think that if AI can produce code for us, human developers
need not learn and practice the principles of clean code. However, there
are two points as counter-arguments to this:

1. AI's results are only as good as the training data it gets. It can
   produce low-quality code that has bugs. The code may appear to be all
   right in spite of being wrong [2]. This will make it hard for humans
   to debug and maintain such code. It is a common observation that it
   is harder to debug and maintain code that someone else has written,
   compared to code that you wrote by yourself.

2. Programming is a creative process. At a very basic level, a
   programmer has to understand the requirements of a real-world
   problem, and develop a solution based on their understanding of the
   underlying mathematics, algorithms, and domain-specific experience.
   Since AI has no inherent understanding of any of these, it cannot be
   expected to produce programs on novel problems. At best it can mimic
   some solutions from its training data.

The above two points imply that although AI can produce code, it is
still not at a stage where it can produce code that is guaranteed to
work. Therefore, human intervention is required at some stage.

Coding is about understanding a problem and implementing it in software.
It will have to continue to be done by humans for some time to come.
Clean coding remains important as long as humans have to deal with code.


2) AI can explain what any code does
------------------------------------

Some others might think that if AI can explain what any piece of code
does, why should a human bother with writing clean code? Whatever is not
understood, can be explained by the AI.

The point to understand here is that AI is fundamentally ignorant of the
purpose of the program.

AI can perhaps describe what a piece of code does. However,
understanding code is not limited to understanding only *what* the code
does. To really understand the code, it is important to grasp the intent
of the programmer. It is necessary to understand *why* a programmer
wrote the code that way. What logic has he implemented?

In the clean code paradigm, the answers to these questions are
communicated by means of properly named objects, and by comments. The
code is written such that it is clear to the reader what the intent of
the programmer was.

Therefore, even if AI can explain what any piece of code does, a
programmer can understand the code better if the code is written well.

It is essential to understand the logic behind a program. That is what
helps to effectively debug, test, and maintain the code in question.
clean code can do that better for us than AI can.


Conclusion
----------

Clean code communicates the logic of the program, and that is what helps
understand and maintain the code. AI is, as of now, not able to produce
perfectly working code, nor is it able to understand the programmer's
intent. Thus, in spite of the ability of AI to help us with programming,
it is still essential for developers to learn writing clean code.


---


References
----------

[1] Robert C. Martin, *Clean Code*

[2] Rodrigo (Twitter:@mathsppblog), *Running Python inside ChatGPT*.
Online: https://mathspp.com/blog/running-python-inside-chatgpt