You’re a computer science student, fresh and new, and you’re being thrown programming exercises to do. Panic on board!
But wait, what’s that thing over there?
Oh yes, it’s ChatGPT, your new best friend! It is able to solve your exercises in no time at all, thanks to natural language processing (NLP) and programming language analysis techniques, but beware, it’s not that simple.
Researchers wanted to dig deeper into the question and see how these AI code generators really influence students’ learning, and for this, they carried out 2 studies. In the first, they took 69 students, complete novices in Python, and separated them into two groups: Those who will use AI and those who will code the old-fashioned way without AI.
For 7 sessions, they gave them exercises to do. The AI-boosted Jedis had access to a Codex-based code generator, a machine-learning model that uses NLP and programming language analysis to generate code from user input. The others had to fend for themselves.
The result of the races?
AI-augmented devs have been a hit! They completed 91% of the tasks compared to 79% for the others. In addition, their code was much more correct. However, on tasks where existing code needed to be modified, the two groups were neck and neck. Then, they gave the students knowledge tests, without the AI. And then, surprise! Both groups had similar scores. But when they retook the tests a week later, the students in the AI-boosted group retained more of what they had learned.
In the second study, the researchers analyzed how the students actually used the code generator. And then, revelations! Some people abused it severely, such as copying and pasting the instructions without thinking. Not cool! 😅 But others were smarter and used it to break down the problem into subtasks or check their own code.
So, what’s the takeaway?
Although AI can be a great tool to learn how to code, it’s only beneficial if you know how to use it intelligently. That’s why tool designers and teachers need to encourage responsible and self-regulated use of these code generators. Otherwise, it’s guaranteed to be dramatic!
As far as I’m concerned, as you know, development is not my thing. But since AI came into my life, “sky is the limit” and it’s helped me tremendously. And like these students, if I put my brain down and go into AI copy-paste mode, in the end, I’m going to have poop. But if I understand what I want to do, if I master my code more like a very technical project manager, well, it’s incredibly efficient. And what’s even cooler is that I learn a lot of stuff. It is often said that you have to practice to become a blacksmith. Well, this is the case, because I don’t bother with syntax issues anymore, and I build my tools brick by brick by understanding everything I do. So AI to develop, yes!! But leaving the brain on.
In any case, one thing is for sure, it’s revolutionizing the way we learn to code. It’s promising for the future, but we will have to make sure that the students really learn to do things and don’t become copy-paste zombies (we already had the problem with StackOverflow, though…).
