N January, Mark Zuckerberg, Meta’s co-founder and CEO, said on the Joe Rogan Experience podcast that artificial intelligence (AI) would begin to take over the roles of mid-level software engineers this year.
Since ChatGPT’s rise in 2022, many new AI tools have appeared, aiming to simplify programming. However, this has also made programmers feel less secure about their jobs, especially those still in school or just starting their careers.
Daren Tan, CEO of Alphv Technologies, believes that learning to code remains important in the AI era, though the purpose has evolved.
Instead of just writing basic functions, coding skills are now crucial for effectively working with AI, customising solutions, and verifying or fixing AI-generated code.
Tan compares the rise of AI to that of calculators: while calculators didn’t eliminate the need to understand math, they changed how we apply it, making us more productive and better equipped to tackle complex problems.
He emphasises the need for a human touch, stressing that “AI-generated code isn’t quite ready for widespread, unsupervised use”.
“While it’s impressive at generating boilerplate code and solving standard problems, it can produce subtle bugs, logic errors or security vulnerabilities.
“It’s best used as a starting point but needs thorough human review and testing, especially for production systems,” he says.
Tan sees AI as playing a supportive role, despite its sophistication, and stresses the need for oversight to ensure everything functions correctly.
“Think of it like having a very knowledgeable junior developer on your team – great at handling routine tasks and offering suggestions, but still needing human oversight for architecture decisions, security considerations, cost optimisations and business logic implementation,” he says.
Based on his experience, Tan says generative AI tools tend to falter when dealing with new or unique situations, especially those that haven’t been encountered before.
Tan sees AI as playing a supportive role, despite its sophistication, and stresses the need for oversight to ensure everything functions correctly. — DAREN TAN
He explained that his team had used generative AI to automate genetic test reports. While the AI was effective at generating basic code for many functions, it struggled with some of the innovative approaches they employed.
“The key is learning to work with AI, not expecting it to do everything. It’s like ‘pair programming’ – the AI can suggest and help, but you as a human need to drive the important decisions,” he says.
Senior lecturer and academic leader at the Asia Pacific University of Technology and Innovation (APU) School of Computing, Au Yit Wah, shares a similar perspective, viewing AI as a complementary tool that serves to make programming more accessible.
“AI is not likely to completely replace human programmers. Although AI significantly changes the way developers work, human programmers are likely to remain essential for the foreseeable future.
“One major reason is the complexity and creativity involved in programming. Developing software often requires solving complex, unique problems that demand deep understanding, critical thinking, and innovation.
“AI, while powerful, still struggles with tasks that involve high levels of abstraction and creative design, such as software architecture and the area of UI/UX,” he says.
UI or user interface refers to the visual elements users interact with on a device, while UX or user experience is about how smooth and efficient that interaction feels.
Au adds that while AI can generate code and even identify bugs in the code when properly prompted via natural language programming, the snippets of code still require human expertise to be integrated into a final software solution.
“Human coders must be able to understand the AI-generated code for refining and integrating it into a larger system.
“Without a strong foundation in coding, a strong logical sense, and knowledge of common programming languages, the human coder will not be able to handle the task of putting the AI-generated code to good use,” he says.
Au highlights that despite the progress made with AI-generated code, reliability and security remain an issue.
“The concepts of secure coding strategies and techniques have not been built into the AI-assisted code generation models. Thus, the code generated by the AI models might demonstrate some degree of security features but is not totally reliable.
“For instance, a study by Stanford University found that a significant portion of AI- generated code had security bugs, underscoring the need for thorough testing and validation.
“Additionally, human oversight is crucial – while AI can assist in generating code, developers must carefully review and test it to ensure it meets security and quality standards,” he says.
Lessons in AI
According to Tan, the area where technology is expected to have a significant impact is in education, as there has been a shift in how coding is taught.
He says syllabuses need to be updated quickly to keep pace with this shift, as there is an increasing gap between what is currently taught and what is needed in the field.
“While traditional CS (computer science) enrolment remains strong, there’s growing interest in hybrid learning paths that combine programming fundamentals with AI tools.
“People aren’t learning less coding – they’re learning differently, focusing more on system design, integration and working alongside AI,” he says.
Institutions like Malaysia’s Multimedia University (MMU) have announced plans to establish a Faculty of AI and Engineering by mid-2025. It is also set to introduce AI components into existing Bachelor’s programmes.
Tan says AI is transforming education by offering interactive debugging assistance, generating practice problems and offering instant feedback.
“However, it’s crucial that we teach students to understand underlying principles rather than just relying on AI suggestions,” Tan says.
Au says that while AI can generate code and identify bugs, human expertise is needed to integrate these snippets into a final solution. — APU
Au similarly believes that there will be a paradigm shift in teaching in the IT and coding space.
“The conventional methods of learning coding, problem-solving skills and creative thinking through programming modules not only need to be maintained but also have to be enhanced.
“This is for the purpose of preparing students to use AI tools at the later stages of their formal education.
Tan highlights the need to practice lifelong learning as a principle when in the tech space.
“I have seen lots of fresh grads struggling when they join tech companies because they’re still stuck with outdated frameworks and concepts that nobody uses anymore.
“The reality is tech moves fast. Like, really fast. What worked two years ago might be completely irrelevant now. Think about it – how many of us were talking about prompt engineering or fine-tuning LLMs (large language models) in 2022?
“The challenge isn’t just updating content according to industry needs. It’s about teaching students how to learn and adapt quickly,” he says.
LLMs are designed to process and generate human language. As a subset of AI focused on natural language processing, they enable machines to understand, interpret, and produce human-like text.
According to Au, while it is hard to say if AI has had an impact on the people choosing a career in coding, AI has made the field of coding more accessible as a whole.
He recommends looking at the situation from two perspectives: first, there are students progressing from secondary to tertiary education. Second, there are working adults seeking career changes and self- improvement.
Both groups are increasingly interested in IT and coding due to the AI boom. However, students typically pursue formal education to learn coding.
This trend has led to a surge in enrolment in IT programmes at higher education institutions, especially in AI-related fields. Conversely, adults looking to change careers often prefer AI-powered learning tools, as these are customisable, more affordable, and have lower barriers to entry.
Au also says these learning tools often come with professional certificates upon completing assessments, which have gained popularity as micro-credential programmes offered through online learning platforms.
Au highlights that despite the progress made with AI- generated code, reliability and security remain an issue. — 123rf
The shifting scenes
Tan has observed a rise in “rightsizing” within the US software development market in recent years, with reports indicating smaller intakes at coding schools and bootcamps. However, this trend has not been mirrored in Malaysia.
Tan reiterates that human programmers will continue to be indispensable despite advancements in AI.
“While AI can handle increasingly complex tasks, programming isn’t just about writing code – it’s about understanding business needs, designing scalable systems, ensuring security, and making cost-effective architectural decisions that require human judgement and accountability,” he says.
He says these skills aren’t going anywhere – they’re becoming more valuable. Often, when dealing with stakeholders or clients, they either don’t know what they want or are unaware of what they don’t know.
“We as technical professionals will need to understand their industry, business rules, daily operations and processes in order to propose the right technological solutions to help them improve or solve their problems or pain points,” he says.
For Au, the key is future-proofing oneself by developing a crucial understanding of how software works, with mastery in traditional programming languages such as Python, Java, JavaScript, and C++.
“Strong problem-solving and analytical skills are also crucial, as coding involves breaking down problems into algorithmic steps and converting them into executable solutions.
“The strong fundamentals in theoretical concepts in computer science and good skills in a few of the commonly used coding languages will ensure your relevancy in the coding profession, regardless of AI.
“This foundational knowledge is essential even when working with AI tools. In addition, learning to code from scratch develops strong problem-solving and logical thinking skills.
“These skills are transferable and highly valuable, regardless of whether you’re writing code yourself or using AI to generate it.
“Having a good foundation in those commonly used programming languages and a good understanding of programming logic and constructs will pave the way for coders to provide more accurate instructions to AI tools to generate what the human coder needs,” he says.
Tan feels the future will be one where AI and programmers are in a symbiotic relationship, where humans can be free to “focus on higher-level problems like system architecture, security and complex business logic”.
“Ultimately, what we want to achieve is AI amplifying human capabilities rather than replacing them,” Tan concludes.
Related stories:
- In Argentina, coding bootcamps target the poor to fill tech jobs
- Should you still learn to code in an AI world?
