ITWeb - From writing code to training AI: A father's perspective

By Johan Steyn, 23 April 2025

Published by ITWeb: https://www.itweb.co.za/article/from-writing-code-to-training-ai-a-fathers-perspective/xnklOvz1k8wq4Ymz

As a primary school student in the late 1980s, I taught myself to code, hunched over an old orange-screen computer, mesmerised by the magic of making machines obey my commands.

By secondary school, I’d built an accounting system for my father’s small business − a clunky but functional program that filled me with pride. Today, however, I rarely write code manually. Artificial intelligence (AI) platforms draft algorithms in seconds, a process which once took me weeks.

This evolution leaves me torn as a parent: my primary-school-aged son now learns programming basics, just as I did. While part of me cherishes this continuity, I can’t ignore the question: in an AI-dominated world, should schools still prioritise coding or pivot entirely to AI literacy and ethical prompting?

The erosion of coding’s competitive edge

My journey mirrors the industry’s trajectory. Early coding taught me problem-solving, but today’s AI tools eclipse those manual processes.

GitHub Copilot, for instance, generates entire code blocks by predicting developers’ intentions − akin to my teenage accounting system, but exponentially faster. A Stanford study confirms this shift: AI-assisted developers complete projects 56% faster, reducing demand for routine coding roles.

The key is teaching coding as a means to understand AI, not as an end in itself.

Yet my son’s curriculum focuses on loops and variables − skills that feel increasingly nostalgic. As McKinsey notes, 45% of coding tasks could be

automated by 2030, disproportionately affecting entry-level roles.

While I value coding’s foundational logic, I wonder: will his generation need to write code − or curate it?

A global educational crossroads

Countries are already rebalancing. The UK’s National AI Strategy for Schools now pairs coding with “critical AI interrogation” skills, teaching pupils to audit AI outputs for bias − a far cry from my 1990s coding isolation.

Similarly, Estonia, once Europe’s coding education poster child, has addedmandatory AI ethics modules, arguing that “debugging society’s algorithms matters more than debugging software”.

This resonates with my experience. My accounting system, while innovative for its time, lacked any ethical safeguards − a gap that today’s students must address as AI permeates finance, healthcare and governance.

Why AI literacy trumps pure coding

My son doesn’t need to replicate my coding journey; he needs to master what I learned too late: AI is a collaborator, not just a tool. AI literacy − understanding model limitations, bias detection and responsible prompting − will define his generation’s competence. For example, when I used AI to automate client reports last year, I spent more time refining prompts and validating outputs than writing actual code.

Schools should mirror this reality. Lessons could involve:

• Analysing how ChatGPT’s training

  data

  skews historical narratives.

• Designing ethical guidelines for facial recognition tools.

• Debugging AI-generated code for security flaws.

  
  

As Unesco reports, these skills cultivate critical thinking and ethical agency − abilities no AI can replicate.

Coding’s niche role: Lessons from my past

This isn’t a rejection of coding entirely. My early experiments taught me computational logic − a foundation for understanding AI’s “black box”. Finland’s hybrid approach recognises this: Students code simple neural networks to grasp AI principles rather than build static websites.

Similarly, my son’s Scratch projects could evolve into prototyping AI-driven solutions. The key is teaching coding as a means to understand AI, not as an end in itself.

Conclusion: Bridging two eras

As someone who coded before the internet existed, I’m torn between nostalgia and pragmatism. My childhood coding fostered resilience and creativity − traits that are still vital in the AI age. But clinging to outdated curricula risks failing the next generation.

The answer lies in synthesis: teach coding’s principles to demystify AI, while prioritising ethical prompting, bias mitigation and human oversight.

As I guide my son through his first Python exercises, I now pair them with discussions about AI’s climate impact and ChatGPT’s hallucinations. This duality − honouring coding’s legacy while embracing AI’s future − offers a blueprint for education in the age of machine collaboration.