Software Engineering is here to stay
Matthew Berman made a bold claim that developer jobs will be obsolete in 10 years.
The word coder appears in Matthew Berman's article sixty-three times in various combinations. Codingjobs jobs have been obsolete for at least thirty years. If you still call yourself a coder, you should be worried -- you are already outdated.
If you are a coder and have coding skills, stop reading here. If you are not already obsolete, you will be in a year -- forget ten years; you don't have that much time. The software industry had no use for coderssince at least the 1970s.
Various predictions about programmer jobs becoming obsolete go back to the 1990s. Ed Yourdon famously wrote a book called "Decline and Fall of the American Programmer" in which he said that outsourcing would replace American programmers. He later followed up with "The Rise and Resurrection of the American Programmer." The reason is that a programmer (or coder) is someone who takes specs verbatim and turns them into software code. A software engineer can see through the ambiguity of user needs and does not need specs to build software.
If you have the imagination, creativity, and ability to build things, you will never be obsolete. If you have neither, then you already are. So trust me when I say this: if you are a software engineer, you will never be obsolete. If you are a programmer or a coder you already are.
Now, let's get back to the original topic. Will AI make software jobs obsolete?
What makes us human is our ability to externalize knowledge and share it. We've been doing this for millennia. We verbally shared knowledge about antelopes by watering holes; we drew things on cave walls and invented alphabets, printing presses, recording devices, books, and art.
Large Language Models are a form of externalized knowledge. They enhance human brain performance, but they don't replace it.
For an LLM to match a human brain's capacity to produce new ideas, the rest of the body would need to go with it. It would need to feel the desires, pains, joys, and suffering humans experience. It would require a sense of community and a desire to share knowledge.
Every line of code written today and placed in a meaningful production environment will require upkeep and maintenance. Most software we use today are not consumer apps. All the transactional systems make the world go around: for example, payments, financial transactions, logistics, and payroll. A transaction processing system processes every purchase you make with a credit card you don't even know exists.
AI is not going to replace transactional systems running deterministic processes. It simply makes no sense for it to play that role. Instead, AI is going to augment the jobs of engineers working on these systems and some aspects of the applications that rely on transactions.
LLMs don't exist in a vacuum. They use APIs to perform useful functions like making payments, ordering goods, and calculating map directions. They ingest information that humans produce.
Apps as we know them today, with UIs and buttons to click, will not be the apps people will want to use in 2-5 years. I am not impressed by over-dramatized demos where someone used an LLM to generate code for a simple social networking website resembling something from the late 1990s.
Asking an LLM to build an app by generating code, posting it to GitHub, and deploying it to Heroku will not be how we build apps using LLMs. There is nothing impressive about LLMs commoditizing basic coding and deployment tasks. You don't need AI for that -- you need automation.
We may get to the point where we can create and share valuable apps without seeing the code behind them. These apps, however, will be nothing like what we use today. They will be a whole new type of app.
Shorthand will still exist. Programming languages are shorthand. When the bubble bursts, we'll all come back to reality in which typing out long, flowery sentences to describe what we want out of our computers is just not an efficient way of using them. A new generation of programming languages and paradigms that include LLM as part of the platform will arise.
Consider what chatGPT does when you ask it to evaluate a mathematical formula: it generates Python code and executes it. Shorthand notation for algorithms has been with us for millennia since the invention of math. Generative AI is not going to replace it.
Final thoughts
I counter the dramatic assertion that developer jobs are on the brink of obsolescence. I distinguish the roles of coders, who may face obsolescence due to their narrow focus on translating specifications into code, and software engineers, whose broad skill set in solving complex problems and innovating ensures their continued relevance. I argue that artificial intelligence and large language models augment rather than replace the human intellect, emphasizing that while app development and deployment methods may evolve, the necessity for software system maintenance and the efficiency of programming languages as a form of shorthand will keep developer roles indispensable. I argue that, despite technological advancements changing the landscape of app development, the core importance of the software engineer's role remains unchanged.