After the development of

After the development of LLM, the developer market and the data analyst market are said to be bad. In particular, the data analyst market is said to be "very" bad, and I think it's for the following reasons.

Corporate data analysts are typically responsible for analyzing business performance. They also investigate the data needed to strategize.

The output of the data analyst is "Analytic Reports".
The analysis report has a "conclusion" and a "base" on this basis.
.

This is the Chain-of-Thoughts (CoT). The reasoning principle of LLM.

So, if the analyst learns from LLM the process of coming to a conclusion, similar results are likely to come out.

If you make a sweat of supervised fine-tuning (SFT) data for CoT learning and strengthen it with the Group Relative Policy Optimization (GRPO), which DeepSeek-R1 revealed as doing well, I think it will produce meaningful results.

If only the data used in the analysis report is automated, the business status analysis will also be automatically performed. Automation does not mean that the performance is bad, but it reflects the powerful performance felt in LLM, so a better quality analysis report can be produced.

You can even use the English-focused model, Llama. Even for a Korean company, the letters are just "Korean," and it doesn't matter that the language is "Korean." Simply put, it doesn't seem to have a problem even if you analyze it in English and translate it into Korean. Llama performs best in English.

It seems that U.S. companies have stopped hiring data analysts in anticipation of this.
.

The Korean legal market is in contrast to this. The legal field also has very structured conclusions and grounds in case law trials, so I think learning with CoT will produce very meaningful results,

Laws should not be translated into English. Both Korean and Korean are important areas. You need Korean LLM to do it properly, but there is no published model. This is why Korean public LLM is definitely needed, even if it is not enough than OpenAI in terms of performance.

In any case, AI is being applied fairly quickly in the legal field, but Korean LLM is needed for in-depth reasoning, so there seems to be quite a hurdle in the process of creating deep and accurate AI.
.

For developers, if you learn GitHub's issues and the PR that deals with them, this will also be highly automated. Structured to look good means that for LLM, it's me.

The developer's future also never seems smooth.
.

As I said in the previous article, AI grows on data. AI is measured and operates within the limits of what it provides as data. Not all areas of the world are measurable.

In the age of AI, it seems that using the five senses will become more important. Up close, soft skills will become more important, and if you look a little further, it seems that the importance of behavioral ability in unstructured situations will increase through various experiences.

This is because the more systematic work and systematic work, the better AI will do.
.

Hundreds of years ago, people directly hammered iron to shape it. However, after the introduction of a machine called casting, the sophistication of casting became impossible for humans to keep up with. This is because casting machines specialize in shaping metal.

AI is also a machine, or a tool, specialized in handling systematic tasks and systematic tasks. Knowing this characteristic will help us live in the future.