Why will programmers know their end?
Discussion on the future of the IT industry and the risks that AI poses to it.
Foreword
I will start by prefacing that this is my personal opinion, based on the experience I have gained as a freelancer in the field, both in subordinate positions and as an "employer".
Although it is a subjective perspective, I believe it accurately reflects the situation the IT sector will face in the coming years.
The reality of AI technology
The AI technology, although not perfect and with evident limitations, represents a reality we cannot ignore. The innovation it brings, though still in its infancy and only a few years old, has already revolutionized the IT industry. Consider that the first viable commercial model emerged only in 2022. And now, in April 2025, we already have AI models that surpass the capabilities of a junior programmer and even more expensive models that can rival an expert, despite the prohibitive costs. currently.
Models like OpenAI's GPT-4 and Google's PaLM 2 have demonstrated impressive abilities in content generation, problem-solving, and even programming. In a very short time, distilled models, smaller in size, have managed to deliver the same quality at costs tens or hundreds of times lower.
The same pattern of innovation has repeated itself. GPT 3 (Davinci) at $60/1M TOK in late 2021 -> GPT 3.5 Turbo (ChatGPT / GPT Instruct) at $2/1M TOK in MAY 2023 -> GPT 4.0 at $90/1M TOK -> GPT 4 Turbo -> GPT 4o and so on.
Even though the price has dropped significantly, the same cannot be said about the ability of AI models to reach the peaks of existing market benchmarks.
I invite you to review some old statistics from late 2022.
I invite you to take a look here as well.
The Risks and Challenges of AI
The emerging problem is the exponential increase in the speed at which AI technology is developing, which brings new risks to the IT field. Risks that, in my opinion, must be taken seriously. First of all, we need to be aware that AI is still a "black box," meaning that as we become increasingly dependent on it, we risk not fully understanding the processes underlying our companies.
The lack of transparency and explainability of AI models is indeed a major challenge. As these systems learn from data in a complex and opaque manner, it can be difficult to understand and explain their decisions and outcomes. This can lead to issues of trust, accountability, and even security risks when we rely too heavily on these opaque technologies.
However, this situation is not without precedent. Take medicine, for example, where many pharmaceuticals we use have mechanisms of action that are only partially understood. Yet, we utilize them because rigorous statistical studies have shown that their benefits outweigh the risks. The same could happen in the IT field. After all, we are talking about two critical domains, and I believe we can agree that a person's life is more important than a software program.
Moreover, companies like Anthropic are making significant strides in explaining the mechanisms that lead to certain completions generated by AI models.
Job Automation and Its Impact on the Market
Companies are increasingly investing in AI, a field that, although currently extremely "overhyped," will be able to automate many tasks that were previously reserved exclusively for humans. We can already use AI to automate menial jobs, for customer support, or to manage tasks that previously required human intervention.
We are witnessing an increasing automation of routine, repetitive, or low-value-added tasks. This has the potential to free the human workforce for more complex, creative, and strategic activities that require problem-solving skills, critical thinking, and empathy. However, it is important to carefully manage this transition to ensure that the benefits are distributed fairly and that no one is left behind.
The issues we are currently observing with AI, such as the lack of understanding of emotions or customer needs, will likely be resolved in the near future as models continue to improve.
The Impact on the IT Market in Romania
In the context of the IT market in Romania, we must consider that we are largely an outsourcing market, which has thrived due to competitive costs compared to more developed countries. However, as AI gains ground, we risk losing that competitive advantage. Many of the tasks that justified the presence of multinationals in Romania can now be automated, reducing the need for cheap labor.
Although there will still be a need for programmers, it is likely that the required number will decrease significantly, especially for junior and mid-level roles, which can be replaced by capable AI models. I have personally experienced how well-rated freelancers have struggled to keep up with the progress made by AIs in just a few months on web development projects.
This should Let this be a wake-up call for the IT industry in Romania. As AI becomes more capable of taking on complex programming tasks, we will need to rethink how we position ourselves in the global market. We must focus on areas where human contribution remains essential - creativity, innovation, leadership, emotional intelligence.
A few responses to anticipated criticisms
In the market, there are several points of critique, some relevant, others emotional:
Pseudointelligence vs intelligence
If something quacks like a duck, flies like a duck, and behaves like a duck, the chances are it’s a duck. Typically, end customers are indifferent to what type of intelligence (or pseudo-intelligence/quasi-intelligence) is behind a program they use.
The rules of competition in predominantly capitalist markets mean that the distinction does not matter if the final outcome is good enough. At certain points, a slightly inferior product may be preferred if it is significantly cheaper.
"The human element, the elusive client"
The main frustration regarding the criticism of technology's potential is the "HUMAN element," as customers are difficult to understand, and therefore, an AI would struggle to comprehend them.
However, I invite you to tell me how often you've cursed at an AI, and to share how many times you've gotten frustrated.
The only thing that AI is not trained to do, due to the way it is trained (not a technology issue), is to respond quickly to avoid boredom. When the focus shifts from instant responses to thoughtful questions, AI models will be able to ask clients and conduct interviews much more deeply and quickly than humans.
The state is being
If you want security, with a little effort, the Romanian state is the perfect parking spot for you, regardless of the emergence of AI.