Titulo Estágio
The software bugs of AI-assisted coding
Áreas de especialidade
Engenharia de Software
Sistemas Inteligentes
Local do Estágio
CISUC
Enquadramento
Artificial intelligence (AI) is set to transform nearly every sector of society, automating tasks traditionally performed by humans. Code development is one of the tasks at the forefront of this revolution, with companies already deploying AI-assisted code (i.e., code written with the assistance of AI) and some even expecting to deploy applications whose code is fully developed by AI from 2025 onwards. Although this may seem like an ambitious target, it is undeniable that AI-assisted coding has become a significant part of current practice in software development. For example, Google reports that over 25% of its new code has been generated by AI, while surveys reveal widespread usage: GitHub found that 92% of developers are "using AI coding tools," and Stack Overflow concluded that over 76% of developers are "using or planning to use AI tools."
The trajectory is clear: AI will play an ever-expanding role in software development. Despite the many advantages of assisted and automatic code development, open research questions must be addressed. Since the advent of software programming, the nature of software bugs produced by humans has been extensively studied. Today, there is a substantial corpus of knowledge on how and why humans produce software bugs. However, AI, like humans, is fallible and capable of generating buggy code that may be deployed into production systems. Given the recency of AI coding tools, there is a significant gap in our understanding of the nature of software bugs produced by AI.
Objetivo
The objective of this dissertation is to perform an introductory study on software bugs found in AI-assisted and AI-generated code and their characteristics.
Plano de Trabalhos - Semestre 1
- T1 (M1-M3): Review of the state-of-the-art in the areas of automated and assisted code generation, LLMs, GPTs and software bugs
- T2 (M4): Choosing set of GPTs that will be used in the study
- T3 (M5): Writing the intermediate report
Plano de Trabalhos - Semestre 2
- T4 (M6): Choosing set of software code and requirements that will be used in study
- T5 (M7-M8): Perform experiments where GPTs are asked to autocomplete or generate code and validate its correctness
- T6 (M9-M10): Writing scientific paper
- T7 (M10-M11): Writing the final report
Condições
The dissertation will take place in the CISUC research centre (SSE group). The dissertation will be co-supervisioned by Prof. Raul Barbosa.
Orientador
Frederico Cerveira
fmduarte@dei.uc.pt 📩