Practical, Learning-Based Tools for Finding and Fixing Bugs

Software bugs are a major problem for software developers and users alike, as they cause crashes, security vulnerabilities, and data loss.

Unfortunately, identifying and fixing software bugs is among the most expensive and time-consuming tasks in software development, accounting for 28% to 50% of the costs of a billion-dollar industry.

The LearnBugs ERC project, on which this proposal is based, has developed ground-breaking techniques to automatically find bugs and to propose suitable bug fixes.

These techniques are based on artificial intelligence and deep learning, making them particularly powerful for kinds of bugs missed by traditional software developer tools.

However, these techniques are currently only available as research prototypes, and there is a gap to be bridged in order to integrate them successfully into the software development workflow.

This Proof of Concept proposal, named BugGPT, aims to make learning-based techniques for finding and fixing software bugs practical and usable by software developers.

The project will develop practical tools that enable software developers to automatically find and fix bugs in their code.

To this end, we will perform technical development activities that address the questions of where, when, and how to suggest bug fixes.

Furthermore, we will perform business development activities to identify potential customers, to evaluate the usefulness of our tools, and to compare potential business models with each other.

Overall, BugGPT has the potential to make a significant impact on the software development industry by making learning-based bug finding and fixing practical for software developers.

If successful, the project could be the beginning of a commercial product that stirs up the market of software development tools.