You’ve seen the headlines. “Yes, ChatGPT Is Coming for Your Office Job.” “ChatGPT Will Replace Programmers within 10 Years.” How serious is the threat? Is it even a threat at all — or is it an opportunity?

Although we cannot know with certainty what the future holds, many writers have expressed very strong beliefs. For example, “ChatGPT Will Replace Programmers within 10 Years” by Adam Hughes reads like the backstory to Terminator 2 or The Matrix.

On the other hand, Torsten Bell argues in a Guardian article entitled “Panic Not. ChatGPT Will Help You Write Better but Won’t Take Your Job — Yet” that in the real world, writing requires organizational and current information that AI doesn’t have, because AI has to be “trained” and only has access to information up through a certain point in time. I will examine the AI threat level from the perspective of two historical jobs, one of which was eliminated by new technology, and the other which survives to this day.

Buggy Whip Makers

The buggy whip maker has become the quintessential example of a profession eliminated by technology. At the risk of beating a dead horse (okay, yes, that was a dad joke), the rise of the automobile eliminated the need for buggy whip makers, along with all other jobs related to carriage manufacturing.

However, a better analogy might be drawn to the auto workers who, ironically, rose in place of the buggy whip makers. When robots started coming for auto workers’ jobs in the 1970s and ’80s, those robots were actually doing the jobs of the humans they displaced, not replacing one industry with another.

This type of technological displacement has been prevalent throughout history. The agricultural and textile industries provide famous examples, such as the looms that replaced textile workers in 19th Century England gave rise to the Luddite movement. Those who predict that AI will eliminate software engineering jobs argue that AI will perform the same work cheaper, faster, and better than humans can.

Lawyers

The legal profession has survived every technological advance in history and, at least in the short term, are widely considered safe from AI as well. Even if ChatGPT could do a lawyer’s job cheaper, faster, and better than a human, the legal field is an example of a field that adopts technology slowly. Several fields, such as banking and the military, are slow to adopt technology for a combination of regulatory and risk-related reasons. These fields require stable, secure, and proven technologies.

The legal profession also involves more than filing paperwork and writing contracts and briefs. Lawyers must also negotiate, argue cases in court, and perform other tasks that cannot be replaced by AI alone. Robotics would also be required, and robots as they exist today are not up to the task.

Something Is Different This Time

Throughout history, the jobs that have been replaced by technology have been those that required the least amount of creativity, discretion, abstract reasoning, and education. Physical, repetitive tasks have always been the easiest to automate. In the case of AI, the new technology has the capability to perform intellectual and creative tasks. Suddenly, it is not so-called low-skilled jobs that are at risk, but office work and creative jobs.

Those who believe AI will eliminate jobs seem to share the belief that software engineering will be first on the chopping block. This argument makes sense. For example, AI can generate original text that seems like creative writing, and the text might even be aesthetically pleasing. However, it is not art; the AI does not understand the meaning of the text it creates. It simply creates a string of words based on probabilities of certain words appearing in a sequence within the text that was used to train the AI. Therefore, the market for creative works by an AI author is limited.

When it comes to software, however, nobody cares about the artistry of the code. They care about whether the software works, how performant it is, and how secure it is. Also, the current methods of creating software are expensive and slow. By contrast, AI can write code in the blink of an eye. For instance, CodingTheSmartWay.com shows how to “Build a Complete Website Using ChatGPT in 1 Minute.” AI is tailor-made to replace software engineers — literally. Therefore, the alarm over the AI threat to software engineering jobs is not unfounded. But is it reasonable?

The Example of E-Learning

I have spent most of my career in the talent development field, in which I have extensively used e-learning authoring tools like Articulate Storyline and Adobe Captivate. Before these rapid authoring tools came along, the development of an e-learning course required a whole team — a project manager, an instructional designer, a graphic designer, and a software engineer. It was an extremely expensive and time-consuming process, which could only be justified when the course would have a massive audience that would otherwise require a huge investment of time, travel, and facility expenses for numerous sessions of an in-person course.

A rapid authoring tool allows an instructional designer to interact with a graphical user interface similar to PowerPoint to develop a course without having to write a single line of code. Like ChatGPT, the tool then writes the necessary code (e.g., HTML). So what happened when rapid authoring tools came out? Did all those project managers, graphic designers, and software engineers lose their jobs?

Perhaps some did. But many organizations still employ such teams to create highly customized e-learning courses for large audiences. For the most part, however, the result was actually the creation of instructional design jobs, because a single person could now create e-learning courses much more quickly. This allowed a much larger pool of organizations to create e-learning courses economically. Project managers and software engineers didn’t lose their jobs, because these organizations didn’t employ project managers and software engineers (to create e-learning courses) in the first place.

The Bottom Line

When the automobile displaced the horse-drawn carriage, buggy whip makers — and most others involved in making carriage parts and carriages themselves — lost their jobs. At the same time, however, the automobile industry created more jobs than it eliminated. When robots replaced factory jobs in the automobile industry (and other industries), they created jobs building, operating, and maintaining robots. Will the same thing happen with software engineers?

In the short term, it seems clear that AI will actually help software engineers by making it easier to learn how to write code and helping skilled software engineers write code more quickly. In its current state, ChatGPT cannot work independently to completely replace software engineers. Someone still needs to make the right requests of the AI (a skill in itself), review the code written by the AI, and indeed program and train the AI itself — even if the AI were to write all the actual code.

The further we try to peer into the future, the less clearly we can see. Only time will tell whether AI evolves to the point that it is capable of completely replacing software engineers and whether we allow it to do so. It is safe to say, however, that before AI can completely replace human software engineers, it would need to overcome several obstacles:

• Industries subject to strict regulatory and security requirements will be reluctant to allow an AI to write their software or to purchase third-party software that was written by an AI.
• Those who predict that AI will replace software engineers envision a homogenous software landscape and code ecosystem. What, then, will differentiate one company’s website or application from another? If, every time a company introduces a new feature, all its competitors introduce the same feature later that day, how can businesses compete? Alternatively, if companies demand complete ownership of their code bases, so that the AI tools used to create those code bases cannot “see” or learn from their own work, then how will those AI tools continue to learn or improve?
• When a bug is discovered in code written by a human, the customer can have a call with an engineer to help the engineer troubleshoot. ChatGPT can’t exactly jump on a call with a customer.
• There is an entire field of sales engineering or solutions engineering. These professionals do not necessarily write code, but they must have coding skills in order to answer customers’ questions and demonstrate their products’ capabilities. This is also not a job for AI.
• Who will program and train each successive generation of AI tools? It might be tempting to conclude that an AI tool could program its own successor, and perhaps it could — but would we allow it to do so? The implications of completely relinquishing control over not only the software that stores our confidential data, empowers our way of life, controls our nuclear missiles, and (in this same future) drives our cars, but also the author of that software, is not something we are likely to do. I would expect legislative and regulatory prohibitions against it.

Is it impossible that AI could, someday, overcome all these obstacles (and any others I’ve missed) to ultimately eliminate all software engineering jobs? Nothing is impossible, given enough time. It does not, however, seem likely. Therefore, the relevant question is probably not, “Will AI replace software engineers,” but rather, “How will AI change software engineering?” We have already seen some answers to that question. I look forward to seeing a fuller answer as AI grows beyond its infancy.