On AI Replacing Programmers

3 minute read

Once enough gray-haired people in Washington DC had installed ChatGPT on their phone, they became collectively conscious that this generative AI thing isn’t a fad. Then they got on the fearmongering bandwagon claiming that artificial general intelligence (AGI) is around the corner (and needs a lot of regulations), and overnight armies of consultants became experts on the topic of how to defend ourselves against this existential threat.

Meanwhile, the NVIDIA stock skyrocketed, Tesla tanked (along with investments in ESG), the US Fed interest rates remained high (along with the USD), and VC firm Andreessen Horowitz raised billions of dollars to invest in AI. Welcome to May of 2024!

Lately, another linear projection entered the echo chambers:

AI is about to replace programmers, so a degree from MIT in Computer Science is worse than it used to be.

Might as well drop out and be the next Zuckerberg equipped with our trusted ChatGPT offering us a competitive edge! ^_^

Sarcasm aside, I think that there is definitely truth to the statement that AI is dramatically changing how we innovate, and that programmers’ jobs are fundamentally changing.

First, it’s true that “anyone can code” now. For instance, have a look at this conversation I had with ChatGPT 4. It’s worth reading it, I promise. ChatGPT 4 is so confident in its abilities! Unlike those pesky programmers, it doesn’t push back and it generates hundreds of lines of code per hour.

Now, do you think it runs? It turns out it doesn’t. It’s too complicated. And today, I will still need a team of developers and the cost to build the prototype for this app will be around 50K for a barebone app.

But fear not, you might say, you just need a bigger context window! So according to this argument, the only real competitive advantage in the world will be the size of the context window available to your organization. I do agree that large context windows are highly valuable.

Yet, this doesn’t mean the fundamental skills of software engineering will become obsolete any time soon. Drawing a parallel with calculators, which didn’t eliminate the need to understand basic arithmetic, the core principles of computer science still hold significant educational and practical value. This ongoing relevance is essential, even as the role of software engineers may evolve towards more oversight and review in light of AI-generated code.

Do you know of a CEO of a Fortune 500 corporation who would accept that the code running their critical infrastructure hasn’t been reviewed by humans? In health care? In finance? In mining? In energy? In construction? I highly doubt that. In fact, we might generate more code now and will no longer need 20M hands-on developers worldwide, but rather 100-200M code-reviewing developers.

The current trend of layoffs and shrinking job openings in tech may be caused by AI advancements but could be influenced by economic strategies and public perceptions. Similar to past shifts in industries like automotive manufacturing, where job dynamics changed due to economic pressures and global competition, the tech industry might be experiencing a redistribution of roles, enabling more global participation and potentially leveling salary scales.

In other words, Google is ok with firing people because Meta is doing it, and vice versa and it has little to do with Generative AI.

In essence, while large language models (LLMs) simplify certain programming tasks, they don’t negate the need for thorough understanding and critical evaluation—skills best cultivated through rigorous study in fields like computer science. This shift towards AI assistance in coding doesn’t diminish the value of a technical education; rather, it highlights the evolving context in which these skills apply.

A good programmer is a smart, patient, and hard working person. People with those traits will continue to be in high demand for a long, long time.