Will AI Replace Developers?

A common question is whether the new “Generative” AI will replace developers in the near future.

This echoes concerns from when computers were first invented—not about replacing developers specifically, but about replacing humans in tedious, mundane jobs. Automation, from the 50s and 60s through the 80s, 90s, and now, has indeed made some roles more efficient, allowing us to do the same tasks with fewer people.

However, the worst-case scenario of everyone being out of a job never happened.

Fast forward to the new millennium, when artificial intelligence emerged in the form of machine learning. The question of replacing developers arose again. Yet, I'm not aware of a single developer losing their job to a machine learning algorithm.

But this new form of artificial intelligence, generative AI or GenAI, is different. It can write code. However, in its current state, GenAI isn't going to take over a programmer's job. For one, GenAI can't participate in meetings or have conversations with human beings about their challenges. GenAI isn't capable of directing such interactions. In short, GenAI currently lacks a real-world interface. It will likely be some time before robotics and other technologies advance to the point where GenAI can direct activities like project teams, business analysts, technical leads, and software engineers do today.

What GenAI is doing immediately is turbocharging software engineers who recognize its potential. The future of GenAI and coding is likely to be more cyborg-like than a Judgment Day scenario. In a sense, we are already cyborgs. From the first time we used a calculator, we became more powerful, doing math quicker and more often, applying it in more situations. As computers evolved, this pattern strengthened. Many argue that while our smartphones aren't embedded in our heads, they are integral to our daily work.

I know of a junior software engineer who didn't complete a software engineering degree but attended a six-month boot camp with a university in Sydney. He went from producing mediocre code to writing at a senior or mid-level engineer's quality in just six months—a process that usually takes five years. The key difference was his embrace of generative AI, specifically ChatGPT. This wasn't a specialized solution, just a generic tool. He used it to write and improve his code, accelerating his learning. He would ask GenAI to review his code and give improvement tips. Over time, he transitioned from low-quality, slow output to high-quality, highly productive work.

I find it interesting that many senior engineers waited the longest to get on the GenAI bandwagon. Even they have realized the power of using GenAI to turbocharge their code. And I'm not talking about a 50% or even 100% increase in output here. In some cases, output has increased five or tenfold, especially in areas new to the developer. Normally, there would be a significant learning curve: taking in new technology, doing some training, writing initial code (often not the best version), and improving it over time. Generative AI shortens that cycle almost to zero by generating the initial code, explaining it, and helping improve it.

In software engineering, there are many repetitive and boring tasks, like documentation and writing unit tests—programs to test other programs. Good unit testing and documentation are crucial for maintaining easy-to-manage code. When picking up an old codebase lacking in these areas, it's tough. Now, when people pick up a new codebase, they first run it through generative AI to explain the code's structure (often different from modern structures and hard to follow), document it, and suggest improvements to the structure and unit tests.

One developer has taken this a step further, running a face-to-face training course for those who have completed boot camps or self-paced learning, helping them get up to speed even faster. Recently, I saw an impressive example: someone with no prior coding experience, part of a project team, was always interested in the technical side. After just three hours of learning how to instruct ChatGPT to write code, she started writing her own code. This helped her understand and challenge the status quo when given timelines and interpreting responses from the engineering team.

So, in the short term, generative AI is not going to replace any software engineers, in my opinion. What it will do is dramatically increase the productivity of open-minded developers, shorten the training cycle for new developers, and lower the cost of building solutions and automations between systems. This is good news for all industries.

Andrew Walker
Technology consulting for charities
https://www.linkedin.com/in/andrew-walker-the-impatient-futurist/