• Schools and training organisations like Simplon and Holberton are adjusting their methods to include training on the appropriate use of AI.
• Students still need to master the basics to ensure that the quality of code is not compromised by an over-reliance on AI.
Whether it be for automating tasks, optimizing and testing code, debugging, data analysis, etc., artificial intelligence is making life so much easier for developers that even Apple is planning on launching its very own rival to GitHub Copilot, the Microsoft coding assistant which has become a reference for web developers. In the long term, many analysts believe that generative artificial intelligence will facilitate the development of code with more support for security that will be better tested and better aligned with rules of compliance. On paper at least, all the indications are that we can look forward to an increase in the overall quality of software, while the work of developers will not require the same level of expertise. However, the reality may well be more complex. Training organizations are already adapting their methods, if only because developers need to acquire new skills to adjust to this new paradigm in the world of coding.
Let’s not forget that AI still cannot produce clean functional code
For Julien Cyr, of the French branch of the international Holberton School of computer science, the use of AI is another natural progression in the history of coding: “Back in the 1970s, coding was done using documentation, lexicons, translators, etc., but over time these tools were gradually superseded by digital resources and online platforms, which simplified the work of development. AI is simply a further step in an ongoing process of simplification.” However, Cyr admits that with generative AI tools: “Students tend to take the easy way out or even cheat rather than trying to understand. But let’s not forget that AI is not capable of writing clean, functional code.”
Demystifying generative AI
For Mathieu Giannecchini, the deputy general manager of another IT school, Simplon, which has trained 30,000 students over the last ten years, the priority is to ensure that teachers are up to speed on the new technology. “We have to be certain that they will not be thrown by students who may be very quick to embrace generative AI for development and infrastructure jobs.” All the students at the school are now enrolled in workshops on generative AI. “The goal is to demystify these tools and to help students understand them better since they will be using them later on in project contexts.” At the same time, Giannecchini believes that students should master the basics before getting to grips with generative AI. “You need to know how to pilot before using GitHub Copilot.” He also points out that the quality of code produced by the new tech is not always up to scratch, for example when it is required to add functionality to an existing application. For coders and training courses specifically focused on development, Simplon goes even further: “We want them to hone their skills with the tools, for example to learn how to use a specific library in an image recognition project.”
Integrating generative AI in students’ working environment
“Generative AI tools are an integral part of our working environment,” points out Julien Cyr. “In short, we recommend that students seek help from ChatGPT before asking technology coaches for assistance. The idea is to encourage them to develop their independence.” Holberton goes even further on its programmes which have only three coaches for every 200 students. In addition to e-learning platforms and peer-to-peer training, the school incentivizes collective effort by offering students from all over the world the opportunity to work on the same projects. And, as you might expect, it has also developed its own automated correction tools to monitor their progress. Finally, as Mathieu Giannecchini is quick to point out, in Web development, what counts is competency. “We cannot blindly put our faith into specific tools.” ChatGPT and GitHub Copilot may well be go-to tools for today’s students, but there is no guarantee that this will always be the case.
Julien Cyr
Mathieu Giannecchini
