Raspberry Pi: Generative AI cannot replace learning to code

Raspberry Pi, a company started with the aim of democratizing computing and recreating the programming frenzy of the 1980s and 1990s, is warning that "vibe coding" cannot replace the skills picked up during the process of learning to code.

In a position paper titled "Why kids still need to learn to code in the age of AI," the company posits that simply shoving plain text into a generative AI and using whatever code it spits out out is not enough.

Rather than running with the output of the likes of GitHub Copilot, the paper argues: "Even in a world where AI can generate code, we will need skilled human programmers who can think critically, solve problems, and make ethical decisions.

"Young people need to learn to code because it is the most effective way for them to develop the mental models and fluency to become skilled human programmers."

Somewhat similarly, there were wizened engineers decades ago who recoiled from newfangled languages like C, insisting that only through learning machine code would new programmers truly understand what a computer was doing.

The situation is a bit different now. If its proponents are to be believed, generative AI is capable of producing working code from a natural language prompt. So why bother with learning the skills necessary to program a computer?

While the Raspberry Pi team acknowledges the benefit of using generative AI output to assist programmers, "skilled human programmers will continue to have a critical role to play in translating messy real-world problems into a form that can be solved through computation."

• AI kept 15-year-old zombie vuln alive, but its time is drawing near
• AI can't replace devs until it understands office politics
• Google crowns Jules to be its agent and spreads the AI love
• AI software development: Productivity revolution or fraught with risk?

"While AI-powered coding tools lower the barrier to anyone being able to generate code, it still takes a skilled programmer to know what good quality, safe, and ethical code looks like.

"The reality is that coding is still the most effective way we know for young people to develop the computational thinking skills that enable them to become an effective programmer."

"Vibe coding," which uses generative AI to produce code from a natural language prompt entered by the user, has become quite the fad over the last few months. The promise of anyone being able to produce functional applications is undoubtedly alluring, but carries considerable risks. AI assistants have a habit of hallucinating, producing plausible but incorrect code, and it is helpful to be able to spot nonsense in the output.

The Raspberry Pi team noted: "The fact that code (either graphical or text-based) does not use natural language as an input is arguably a feature and not a bug.

"The friction introduced in the conversion of human reasoning into a rigid expression of logic is where the learning and development of computational thinking occurs." ®