Are tools like GitHub Copilot the future? AI in software engineering

A topic that no one has been able to avoid in recent months? Artificial intelligence (AI).
Whether for process optimization in companies, as chatbots like ChatGPT or legally questionable AI apps: “For only 4.99 euros, we create avatars that look a bit like you at best” – we remember.

One of the truly groundbreaking AI innovations is Copilot, a coding assistant developed by GitHub and OpenAI. Copilot aims to change the way software engineers work by making intelligent suggestions and automating repetitive tasks.

In this blog post, we look at what the tool can do, how we at BAYOOTEC program with AI in software engineering and consider the opportunities and risks. In the long term, can such programs really help us to make engineering projects even more efficient? Or will AI replace our jobs in the future?

In short, GitHub Copilot works on the basis of machine learning and is aimed specifically at software engineers to assist with programming. Step by step, the tool analyzes code that has already been written and thus learns about structures and patterns. Based on these, CoPilot then suggests suitable code segments – a great time saver.

The program uses not only its own code, but also a variety of data sources such as publicly available repositories on GitHub. Copilot is thus capable of solving even complex tasks.

It also offers a number of features such as the ability to create control flow diagrams, avoid conflicts between codes, and automatically check the quality of the code.

Since the advent of Copilot, the BAYOOTEC team has realized how much time AI tools can save in software engineering. Data-driven applications are what we focus on in many projects. Often, however, we need test data for the project that is as close as possible to reality, data that we can use for engineering.

A small example: For organization X we create a data-driven application to optimize their digital processes and save costs. Organization X provides us with 3 or 4 sample records, including names and email addresses. To engineer the user interface and frontend, as well as to be able to detect performance problems in advance, we need 10,000, if not 100,000, data sets.

“Nothing easier than that,” think readers of this blog, “you can copy the data sets.” You could, but it is not useful because the data sets do not cover all variants. Of course, we could write a small program ourselves, but Copilot solves this problem for us, keyword workload reduction.

Through pattern recognition, the tool recognizes names in the sample dataset and creates 10,000 sample names for us. It recognizes the email address and creates 10,000 more sample addresses.
In the user interface, for example, we notice that the default characters are not sufficient for all variations of names or that we have not allowed hyphens for “Kai-Uwe”. As software engineers, we benefit from this because our code is better and our customers benefit because our code contains fewer potential problems.

Another important point is understanding code that is not written by us. We often work with legacy code that we receive from customers. It’s not always of the best quality – after all, there’s a reason why we, the experts, are now working on this code rather than those who originally created it. But before we go through the hassle of trying to figure out ourselves what the author was trying to tell us, Copilot can quickly and easily explain even code that is not the best in terms of quality.

Additionally, Copilot has a good nose when it comes to detecting potential bugs. It’s about the non-obvious bugs (of course, our engineers don’t create obvious bugs), it’s about the kind of bugs that only come out of context. Copilot reliably detects these and points them out. Sounds like the omniscient AI tool is so powerful that it could do all the work on its own right away, right? So why do we still need software engineers at all?

Even beyond AI tools, artificial intelligence is part of our everyday work. With AI, we optimize processes in organizations, increase security and reduce costs, especially in the field of e-commerce. How does it work? You can read about this here in our BAYOOTEC service overview.

For all the advantages that Copilot offers, there is one decisive shortcoming that will save us all our jobs: there still needs to be a person behind the program who knows which functions should be implemented in the first place, who understands and masters the big picture.

Although many people fear (as with most innovative technologies) that AI will replace jobs in the future, AI in software engineering should be seen much more as a hard-working helper that assists software engineers and saves time that would otherwise be needed for manual tasks.

AI in software engineering will not replace our jobs, on the contrary: by using AI to make our work easier, we at BAYOOTEC have the opportunity to invest more time in creating innovative ideas and features. Ultimately, this also helps our customers, whose software solutions we continuously engineer and improve.

In a nutshell: GitHub Copilot is a brilliant approach and an important step towards integrating AI tools into software engineering and making manual processes easier for developers.
Of course, neither Copilot nor any other AI will be able to cover every aspect of software engineering – but that’s not what the tools are designed for. Human creativity, experience and expertise are irreplaceable, but AI tools can definitely already help us work more productively and efficiently. And considering the possibilities AI has already brought in other areas, we should by no means close our minds to its ability.

All in all, we can look forward to an exciting future with AI-based tools. And who knows, maybe they will open up possibilities for us that we are not even thinking about today.