AI doesn't retire programmers; it turns them into code reviewers

AI doesn't retire programmers; it turns them into code reviewers

A higher vocational training student in DAM shares on a forum: when pressed, he uses ChatGPT to get programming exercises for services and processes in Java, adjusts the final code and moves on. He tells it without guilt or epicness. His doubt isn't whether the tool works, but what happens to the profession when everyone does the same. The response received in the forum isn't unanimous: some argue that AI is a mediocre corrector that hallucinates as soon as you go beyond the basics, and others believe that the programmer as we know him has an expiration date.

Does AI replace the programmer or just change his job?

The first line of fire is occupied by those who already use it daily. A participant with years of experience says that he passes an input example and what he wants out of output, well explained, and usually does quite well. Even asks to translate code from C# to ABAP and most of the times it works. Another points out that with GPT-4, Java increasingly catches more errors as it learns the language, and that it doesn't have creativity based on human experiences. Translation: it serves for mechanical things, not for what requires criteria.

The underlying problem is that the tool deceives. It never says "I don't know." If it doesn't know an answer, it invents it, even if it's made up reserved words that don't exist and code that doesn't compile. For learning it is terrible, because directly you don't know if what it is giving you works. A participant checked it with a multiple-choice test from a occupational health and safety course: a 6 out of 10 on questions at the level of not putting fingers in the plug.

The competitive advantage isn't in typing code, it's in knowing what to ask. Several messages agree that the future of the programmer is as intermediary between the client and the machine: the person who understands what has to be done and how to ask it to the AI. Until, within a few years, it will be the client who asks directly because the AI will already know how to interrogate him. That scenario doesn't affect only computer science: doctors, lawyers and civil servants enter in the same sack.

The water carrier argument: technology creates jobs nobody imagines

The most cited reply is historical. When running water arrived, the water carriers who carried water to houses were left without a trade. How many jobs did running water generate? Treatment plants, reservoirs, pipelines, plumbing, pumps, maintenance, taps, administrative management, wastewater treatment. The conclusion drawn is that you have to be very short-sighted not to see the avalanche of posts generated by each technological leap.

The parallelism is completed with photography: purists said that where there is a Hasselblad, out with those digital toys. We already know who won. And with WordPress, which took away much work from programmers, but someone had to program and maintain it. The question that closes the circle: who programs and maintains AI?

There is a data that disconcerts the most optimistic. A participant recalls that a company has been using an AI in customer service for nearly three years and it is almost impossible to notice. If that occurs at the first step, the trinc models will have capacity to substitute those above. Except for executives, who are untouchable.

Manual trades, the refuge AI cannot reach

The most transversal consensus of the exchange isn't in programming, but in drywall. A message summarizes the exit: stop programming and learn to install drywall, something AI cannot do. No AI tiles a bathroom. A construction worker develops it in detail: a robot wouldn't know how to access the site, nor which floor to go to, nor which room, nor press the freight elevator button, nor unload a truck. It couldn't do anything by itself beyond plugging the cable in a specially designed laboratory.

The repeated recommendation is to have knowledge of another field where to apply programming. A biochemist who also programs can create tools that no programmer without his knowledge would make. The creative role based on direct experience is what the machine doesn't equal: a participant suggests that an AI wouldn't occur to design tennis technique correction software with augmented reality glasses and GPS watch if it never played.

Vocational training, the degree and the lies told

The initial message also points to another discussion: it isn't true that vocational training is worth more than university. No HR person will value a vocational graduate the same as a university one, he argues, although the phrase is cut off in the available material. What does appear closed is the warning from an industry veteran to new generations: don't get into this even drunk, at least in Spain. Twenty or thirty years ago it was worth it, but now every few months a new version of the framework comes out and you have to update.

The most pessimistic calculation gives ten years margin. It is the deadline a participant grants before things really change. Another summarizes it: for now you have a job, AI is a tool to make you more efficient. And a third adds the nuance nobody disputes: AI is too smart to want your job.

The exact point where the analysis gets stuck is always the same. Nobody knows if the programmer of 2035 will be the one writing the code or the one explaining to the client why the machine invented everything with impeccable confidence.