How to check if code was written by AI?

How to Check if Code was Written by AI?

The use of Artificial Intelligence (AI) in software development has become increasingly prevalent, with many developers using AI-generated code to speed up their work. However, it’s crucial to verify whether the code was indeed written by a human or AI. As a developer, you want to ensure that you’re working with high-quality code that meets your standards and is maintainable. In this article, we’ll explore the ways to check if code was written by AI and provide guidance on how to verify its authenticity.

Why Check if Code was Written by AI?

Before we dive into the methods of checking, it’s essential to understand why verifying the origin of the code is crucial. AI-generated code can have several limitations, such as:

• Lack of contextual understanding: AI algorithms may not fully comprehend the context and nuances of the problem, leading to suboptimal solutions.
• Coding style and conventions: AI-generated code may not adhere to coding standards, making it harder to maintain and update.
• Security risks: AI-generated code may introduce vulnerabilities, compromising the security of your system.
• Lack of human touch: AI-generated code may lack the human empathy and creativity that’s essential for building user-centered applications.

Methods to Check if Code was Written by AI

Here are the methods to verify if code was written by AI:

1. Manual Code Review

• Visual Inspection: Look for unusual code patterns, such as repetitive syntax errors or inconsistent naming conventions.
• Code Style: Check if the code follows standard coding practices, such as indentation, spacing, and naming conventions.
• Indentation and Spacing: Look for non-standard indentation, excessive spacing, or inconsistent formatting.

Code Example:

if ( condition ) {

    // code block

    if ( anotherCondition ) {

        // another code block

    }

}

2. Code Analysis Tools

• Code Duplication Detection: Tools like duplication detectors (e.g., duploc) can identify duplicated code sections.
• Code Metrics Analysis: Analyze code metrics, such as cyclomatic complexity, Halstead’s complexity, and Maintainability Index, which can indicate if the code is overly complex or fragmented.
• Code Analysis Tools: Utilize tools like SonarQube, CodeHeat, or CodeCoverage to analyze code quality, security, and maintainability.

Code Analysis Example:

// Code with high cyclomatic complexity

if ( condition ) {

    if ( anotherCondition ) {

        if ( thirdCondition ) {

            // ...

        }

    }

}

3. (pathological) Code Structure

• Repetitive Patterns: Look for repetitive patterns, such as:

  • Circular dependencies
  • Deep callbacks
  • Unnecessary complexity

• Modern Programming Paradigms: Check if the code uses modern programming paradigms like functional programming or aspect-oriented programming.
• Legacy Code: Identify if the code is using outdated languages or frameworks.

Pathological Code Example:

// Circular dependency

from a import A

from a import B



def a():

    if do_something_with(B()):

        return a()

    else:

        return something_else_with(A())



def B():

    # ...

4. Linguistic Analysis

• Natural Language Processing (NLP) Tools: Utilize NLP tools to analyze the code’s natural language patterns, such as:

  • Sentiment Analysis
  • Named Entity Recognition
  • Part-of-Speech Tagging

• Coding Comments and Documentation: Check if the code has proper comments, documentation, and use of English language conventions.

Linguistic Analysis Example:

# This function calculates the sum of two numbers

def add(a, b):

    return a + b

5. Code Obfuscation Detection

• Code Obfuscation Detection Tools: Utilize tools that detect code obfuscation, such as Adding numbers or using Reflection.
• Specially Designed Patterns: Look for patterns, like using uncommon or dialect-specific code styles, to identify potential AI-generated code.

Code Obfuscation Example:

package obsolete;

public class OverloadedClass.java {

    public void method() {

        return (byte) (0xFF & (int) new String(((Object)o).toCharArray()));

    }

}

Conclusion

Verifying if code was written by AI is crucial to ensure the quality, security, and maintainability of the software. By utilizing the methods outlined in this article, you can identify potential AI-generated code and take corrective action. Remember that AI-generated code may not always be perfect, and human expertise is still necessary to refine and improve the code, as well as to provide context, empathy, and creativity.