The Over-Hyped AI These Days

Recently, there has been a trend of attaching AI to everything, everywhere. There are even talks of an AI toothbrush coming soon, which gives off the feeling that it’s being attached mindlessly for marketing purposes.

A few days ago, I saw a flyer in front of my house that read, "AI Math!" I had no idea what that meant, and it didn't resonate with me what kind of value it aimed to provide to consumers.

The problem is that just attaching AI to everything often leads to using inadequate AI models, which I believe deviates from the essence of what AI should offer. It boils down to a mentality of "It looks impressive!"

Conversely, I fully support utilizing AI for genuine value creation.

Developers in the Age of AI?

The trend is that many developers are being replaced due to AI. In fact, I got my first job as a developer in 2021, when there were lots of government-funded training programs flooding the market with developers.

Now it feels like half of them have been wiped out, as if facing Thanos, and the government-funded classes have nearly gone extinct.

Worrisome Trend?

Senior developers do not just fall from the sky; they grow into that role through juniors continuously making mistakes, enduring hardships, and enduring long formative periods.

The issue is that many companies are replacing juniors with AI for immediate productivity. Of course, I am not saying to keep juniors who lack skills. If there is no willingness to learn and their learning ability is poor, it is the same in any field for them to be eliminated. The problem is the one-dimensional thought of "Isn’t it fine if AI does everything? It’ll be faster if a senior operates AI!"

Moreover, AI learns mindlessly from a multitude of codes. It does not understand which code is quality. Therefore, if one blindly follows the code used by AI, they will use the "common" codes, rather than the "best" code for their situation and context. I think the Pareto Principle applies in almost everything; the "majority" result cannot be the "best".

And just like Garbage in, Garbage out, if I do not know, I will ask AI only to that level and accept that code. Personally, I greatly dislike vibe coding because it seems irresponsible to rely on "intuition". A critical mindset exploring "Is this the best method?" is essential to problem-solving, while just copying and memorizing from answers actually degenerates the "problem-solving ability" that developers should pursue.

Then Should We Not Use AI?

It's similar to telling mathematicians not to use calculators or office workers not to use Excel. It’s not that you should never use them. Rather, it’s about developing the habit of thinking for yourself before asking AI. Instead of saying, "Please model this for me; I want XX," you should say, "I’ve created this model for XX; what do you think?"

Prioritize search engines and official documentation. The best quality code and processing methods are in the official documentation and a well-written repository by someone.

I believe simple repetitive tasks (translation, writing test codes, and stylistic adjustments) can greatly reduce my time thanks to AI. However, for juniors to immediately seek answers from AI during the design or business logic writing process is toxic.

Conclusion?

Use AI. It’s fine to do so. However, you should strive not to become a developer who cannot code without AI. Additionally, researching to gain a solid understanding of CS and low-level concepts will serve as a foundation for new problem-solving.

Focusing on creating a "good" service rather than just any service is essential.