Cloud Migration of an Outdated On-Premises Web Application to an AWS Hosting Solution; An Ethical Analysis of GitHub Copilot from A Utilitarian Perspective

Author:
Glance, Franklin, School of Engineering and Applied Science, University of Virginia
Advisors:
Laugelli, Benjamin, EN-Engineering and Society, University of Virginia
Krzysztofowicz, Roman, EN-SIE, University of Virginia
Abstract:

My technical capstone project and STS research paper explore related topics in the field of computer science and artificial intelligence. In the technical report, I detail the migration of an outdated distribution fitting application ‘DFit’ to the cloud. This project involved careful consideration of security and performance aspects of the new architecture when upgrading DFit to ensure cloud compatibility. My STS paper examined the ethical implications of GitHub’s decision in releasing Copilot. In the paper I argue that GitHub acted morally in developing and releasing copilot when examined with the ethical framework of utilitarianism, since the benefits of improved developer productivity outweigh the concerns surrounding potential copyright infringement.
For my technical project, I migrated a distribution fitting web application from an on- premises hosting solution to the cloud. This task was part of my Systems Engineering capstone project. The existing codebase’s inadequacies in performance necessitated a transition to a modern hosting environment. This transition required a technical redesign of the entire system due to the outdated technology used in the on-premises hosting solution. I developed a comprehensive redesign for the website that maintained the functionality of the old version while markedly improving scalability and performance using Amazon Web Services (AWS) hosting and the Django web framework. This project demonstrated the vast benefits migrating to cloud hosting can have for a web application and provided useful experience with working on a real- world software engineering problem.
In my STS research paper, I examine the ethical concerns surrounding GitHub Copilot. Copilot is a coding assistant that uses a Large Language Models (LLMs) to provide coding suggestions during the development process. The LLM backing Copilot was trained on public
code hosted on GitHub, which introduces potential for copyright infringement. The copyright concerns relate to the potential for generation of code nearly identical to copyrighted training material. Using the ethical framework of utilitarianism, I argue that GitHub acted morally in developing and releasing GitHub Copilot because the benefits it provides in terms of improved developer productivity and innovation justify the potential drawbacks related to copyright concerns. In this analysis, I use the utilitarian concepts of the utility principle and harm principle to investigate the morality of GitHub’s decisions.
Working simultaneously on both projects during my fourth year has provided valuable insights into the interaction between software engineering and ethical considerations that must be made when developing AI powered technologies. The DFit migration shows how rebuilding old software can vastly improve its safety and performance, and the Copilot analysis highlights the importance of critically examining the net impact of developing and releasing new software. The technical project helped inform the STS research paper by providing a concrete example of how effective programming can vastly improve existing technology. Insights from the STS paper guided the ethical considerations made when developing the new version of DFit. The technical and ethical experience gained from working on these projects will shape how I approach developing and deploying AI-powered software in my future as a computer scientist.

Degree:
BS (Bachelor of Science)
Notes:

School of Engineering and Applied Science
Bachelor of Science in Systems Engineering

Technical Advisor: Roman Krzysztofowicz
STS Advisor: Benjamin Laugelli
Technical Team Members: Esther Yi, Mitch Mitchell

Language:
English
Rights:
All rights reserved (no additional license for public reuse)
Issued Date:
2024/05/02