Raspberry Pi, a company that aims to democratize programming frenzy and replicate programming in the 1980s and 1990s, warns that “Vibe Coding” cannot replace the skills featured during the process of learning code.
In a position paper titled “Why should kids learn to code in the age of AI,” the company assumes that simply pushing simple text into generative AI and using spitting code is not enough.
Rather than running it on output such as Github Copilot, this paper argues:
“Young people need to learn to coding as it is the most effective way to develop mental models and urgency to become skilled human programmers.”
A little likewise, there were wise engineers who reacted to a new language like C decades ago, claiming that only through learning machine code new programmers would really understand what a computer is doing.
The situation is a little different now. If its supporters are believed, generative AI can create working code from a natural language prompt. So why are you troubled by learning the skills necessary to program a computer?
The Raspberry PI team acknowledges the benefits of using generated AI outputs to assist programmers, but “skilled human programmers continue to play a key role in translating troublesome real-world problems into formats that can be solved through calculations.”
“Ai-powered coding tools reduce barriers to those who can generate code, but to skilled programmers you need to know what quality, safety and ethical code looks like.
“In reality, coding is the most effective way to develop computational thinking skills that allow young people to become effective programmers.”
“Vibe Coding,” which uses generated AI to create code from user-entered natural language prompts, has become quite popular in recent months. The promise that anyone can create functional applications is undoubtedly appealing, but it does have considerable risk. AI assistants are helpful when they have hallucination habits, generate plausible but incorrect code and can find nonsense in the output.
The Raspberry PI team states, “The fact that the code (graphic or text-based) does not use natural language is definitely a feature as input and not a bug.
“The friction introduced by transforming human reasoning into strict expressions of logic is where learning and development of computational thinking occurs.” ®
