Next-Gen Software Engineering. Big Models for AI-Augmented Model-Driven Software Engineering (2409.18048v2)

Abstract: The effectiveness of model-driven software engineering (MDSE) has been successfully demonstrated in the context of complex software; however, it has not been widely adopted due to the requisite efforts associated with model development and maintenance, as well as the specific modelling competencies required for MDSE. Concurrently, AI methods, particularly deep learning methods, have demonstrated considerable abilities when applied to the huge code bases accessible on open-source coding platforms. The so-called big code provides the basis for significant advances in empirical software engineering, as well as in the automation of coding processes and improvements in software quality with the use of AI. The objective of this paper is to facilitate a synthesis between these two significant domains of software engineering (SE), namely models and AI in SE. The paper provides an overview of the current state of AI-augmented software engineering and develops a corresponding taxonomy, AI4SE. In light of the aforementioned considerations, a vision of AI-assisted Big Models in SE is put forth, with the aim of capitalising on the advantages inherent to both approaches in the context of software development. Finally, the new paradigm of pair modelling in MDSE is proposed.

• The paper demonstrates how integrating AI with MDSE can address complex software challenges by leveraging Big Models and extensive code data.
• It details novel techniques for extracting and transforming models from large code repositories and introduces an updated AI4SE taxonomy.
• It proposes the innovative pair modelling paradigm where engineers collaborate with AI tools to improve code quality and strategic development oversight.

AI-Assisted Big Models in Software Engineering: An Insightful Overview

The paper, "Next-Gen Software Engineering: AI-Assisted Big Models" by Ina K. Schieferdecker explores the integration of AI techniques, particularly ML, with model-driven software engineering (MDSE) to address the increased complexity and diversity of modern software systems. This synthesis aims to capitalize on the complementary strengths of both domains to enhance the software engineering process.

Background and Issues in Software Engineering

Software engineering (SE) strives to establish principles and best practices for reliable, secure, and scalable software development. However, traditional practices have not kept pace with the quality demands of critical domains. This is illustrated with examples like the significant error density in large-scale software projects and the unprofessional software development practices highlighted by recent outages.

MDSE has demonstrated efficacy in handling complex software projects, although its widespread adoption is hampered by the efforts required in model development and the specialized skills needed. On a parallel path, AI techniques, including ML, have made substantial progress, particularly through the availability of extensive code bases on platforms like GitHub. The convergence of AI and MDSE in SE could address numerous existing challenges by leveraging Big Models, analogous to the concept of Big Data.

Models in Software Engineering

Models are central to managing the complexity of software through abstraction, supporting design decisions, and providing comprehensive overviews of software structures and behaviors. MDSE has evolved beyond traditional top-down approaches to include bottom-up model extraction from software executions. This dual-directional approach, incorporating data structures often seen in JSON or XML, forms a foundation for integrating AI into SE.

Big Code in Software Engineering

The prominence of open-source platforms like GitHub has led to the concept of Big Code, characterized by vast repositories of software code. Analysis and reuse of this Big Code have advanced empirical software engineering and allowed for training ML models to further automate coding processes. Platforms like Kibble facilitate insights into coding projects, revealing that a significant portion of code consists of information structures.

AI in Software Engineering

The utilization of AI in SE spans various use cases, from enhancing code understanding and generation to improving software quality. The paper introduces an updated taxonomy for AI in SE (AI4SE), categorizing applications into understanding, generation, and improvement goals. These applications span multiple phases of the software lifecycle, from requirements to maintenance, and extend to SE process management and operations.

AI-Assisted Big Models

The integration of MDSE with AI, leveraging Big Models, is portrayed as an emerging next-generation SE practice. Key opportunities include:

1. Providing extensive top-down models on coding platforms.
2. Extracting models from Big Code.
3. Forming Big Models to advance empirical MDSE.
4. Applying AI methods to Big Models.
5. Establishing the paradigm of pair modelling.

Recent studies support these opportunities, such as the SEMI Software Engineering Models Index for repositories and the Lindholmen dataset of UML models. Techniques extracting models from code and execution traces are instrumental in maintaining alignment with current software states. The potential of AI applications in Big Models is explored with promising initial results in classification, model merging, and transformation learning.

Pair Modelling Paradigm

Pair modelling emerges as a natural evolution from pair programming. It involves a collaborative approach where a software engineer and an AI tool alternate roles as driver and observer/navigator. This approach enhances software artefact development, ensuring both immediate development focus and strategic oversight. The efficacy of this paradigm depends on the sophistication of AI tools and their levels of support according to the AI4SE taxonomy.

Conclusion

This paper highlights the significant potential of combining AI with MDSE to advance software engineering. The AI4SE taxonomy provides a comprehensive framework to categorize AI applications across the SE lifecycle. The concept of Big Models, supported by AI-driven pair modelling, represents a promising direction for future research and practice. Ongoing work to refine these models and frameworks will further elucidate the transformative impact of AI-assisted Big Models in software engineering.