How much longer will the manufactories last? / Hebrew
Before reading the article, I would like to clarify two things:
1. It should not be regarded as an expert opinion, rather as a private individual working in the industry
2. This is not a prediction, just a guess, based on reading materials, social networks and news. It is rather an attempt to understand what is happening, to explain it first of all to oneself.
For the first time, this feeling overtook me, probably like many, in 2023, when ChatGPT from OpenAI suddenly shot up. At that time, the professional infofield was already raging and the market was beginning to be filled with all kinds of “neurosets”. The feeling that visited me was anxiety, the thought that the moment will come when some conditional Copilot will do all the work for you.
According to my feelings, designers and artists were the first to feel the impact, when any picture is drawn in a matter of seconds, while a professional artist needs several hours (minutes at best).
In addition to anxiety, there was, of course, a second reaction. It arose after the first interaction with ChatGPT, when I needed to write a small bash script, and I did not want to spend a long time with it. What was my surprise when the generated script worked literally after one or two edits that were purely related to my specifics. Then I thought that it greatly relieves time and allows me not to be distracted by routine (writing test scripts, creating simple automation). Now you can focus purely on coding. At that moment, I felt empowered because such a tool expands my capabilities as a programmer. However, over time, while reading the news in a hurry, more and more dissolved.
And that’s when an interesting thought came to mind – an event took place in the industry, the scale of which resembles the first industrial revolution.
The first industrial revolution
In short, there was a transition from manual labor to mechanized labor, from a manufactory to a factory. A very vivid example: weavers and weavers, who made cloth by hand, spending a lot of time on it, were replaced by looms, which accelerated production, increased the quantity and lowered the price of cloth. A previously profitable profession during the transition to factories simply became unnecessary. Home weavers and factories went bankrupt as a result.
It is worth noting that since it was necessary to build and adjust new factories in the changing realities, the skills of maintenance of weaving machines, their adjustment, design and repair became in demand. Having destroyed one profession, the industrial revolution gave life to others that require more skills and, at a minimum, general literacy. Because the craftsman could verbally explain to the younger generation how to weave correctly. When working in a factory required a person to at least be able to read in order to understand the instructions for working with the machine.
Everything is new – the old is well forgotten
A similar process, in my opinion, is taking place right now.
Before the widespread introduction of language models, all IT companies were factories from the point of view of the production process: people sit, write code, create some products. Everything is done by hand. There is a lot of duplication, a lot of identical pieces of code that differ from each other in details. I don’t want to say it’s bad. This is reality. In these considerations, we will omit the level of training, because it varies greatly depending on the tasks. Somewhere such solutions are being developed and applied that no one has ever done before. And somewhere – completely banal products that use elements of others. Within the framework of this reasoning, we note only that everything is invented and done by hand. Many will notice: – what about testing and debugging, drawing up projects? I will only answer that all this is auxiliary, that’s all helps create a product, but does not create.
The use of language models shows that their code generation is faster than that of an ordinary developer. And not average either. Currently, the quality of this code generation leaves much to be desired, because the “neural network” can often “invent” an element that does not exist in the project or library. “Neuroset” may make a mistake in the implementation of the algorithm and it will have to be debugged. But these are only details. If the amount of generated code is already impressive, then the quality is only a matter of time. Sooner or later, the quality will increase so much that there will be no need to edit not only syntactic structures, but even the logic of the project itself. The most important thing in this case is to set the parameters for code generation correctly: what the user interface elements should look like and what properties they should have, what the load is based on the number of users, what tasks the project should perform – video streaming, game server, etc.
Thus, we come to the conclusion that the main skill will be the ability to work with “neurosets” – to teach them, ask questions correctly and give hints that lead to a result as close as possible to the required one. That is, you will need to be able to configure the machine that generates the working code according to the specified requirements, with the specified properties, in the shortest possible time. If something does not meet the requirements, adjust some parameters through the same requests so that the machine rewrites part of the project and replaces it with a block that does not suit us.
Some obscure professions
I remember laughing for a long time when I saw the job description “Prompt engineer”. Now, reading posts about how schoolchildren skillfully compose prompts for bots to help with the house, and how cleverly you can remove ethical restrictions from a chatbot by simply saying that you need to help a cat stuck in the basement, you begin to relate to the emerging field II to new professions, if not more seriously, then already without the former smile. There was approximately the same reaction to the name “AI-trainer”.
In our rapidly changing world, it is very important to keep your finger on the pulse. Events show that classical programming will lose its former popularity, but basic knowledge will remain relevant: OS device, basic algorithms and data types, basic understanding of the device of various “neural networks”. Programming itself, as an activity, will most likely remain in some specific areas. The only thing that can remain more or less unchanged is system and network administration skills. Even if assistants will be built into the OS (not in the kernel itself, but as a system on top of it) and network configuration systems, the skills to manage them and configure them will still be in demand.
I cannot say that the changes taking place in the industry cause panic.
So far, there have been no drastic changes, but some news makes you not just think, but already take the first and confident steps to not be left behind.
The question in the title of the article remains open. In my opinion, only time will tell how much time is left for the classic manufactories of programmers. But one thing is clear: all companies that want to stay afloat will switch to code generation or are already doing it.
Recently, after watching the movie “Hidden Figures”, I came across an episode that in some ways resembles the current situation: the head of the computer department (in English – computer) at NASA saw how a computer was being installed in one of the offices, which disturbed her. Then she found out that computers were being installed in the enterprise so that her department would be completely replaced. The woman understands in time that if she and her employees do not learn the programming language, then everyone will be out of a job.
Attention, spoiler ahead!
At the end of the film, of course, there is a happy ending – under the sensitive leadership of this woman, the computers in the entire department are retrained as programmers.