Getting_tired_of_ai

Well, it is probably both exciting and mentally exhausting to live these days as a person who does programming for a hobby and for a living.

It is exciting because with a fraction of your time and physical efforts (aka typing), you can materialise any systems that you have in your brain regardless of how complex it is (as long as you can afford the API costs of running the llm or pay an expensive subscription to cover your usage).

Although one caveat is that you should be pretty good at describing and articulating what you actually want to materialise or get answers to. Otherwise you are hoping to hit the bull’s eye with your eyes closed. Imagine, on a random casual day, your boss comes at your desk and start describing this new idea that they think is cool but with vague and abstract words. You, being asked to materialise this from a brief conversation, would be completely lost.

This is where the mentally exhausting part comes in. The real world problem arises when there are any stakeholders that think AI is some kind of a Harry Potter’s magic wand and they pass on the burden of thinking, planning, and conceptualising to AI completely. When it comes to simple prototyping of idea, this actually works and probably that is how it should be done. However, when it comes to maturation and productionisation of the idea, this way of working with AI will unlikely get you to the final destination. Maybe if you are lucky and a god of AI draws the lucky dice and hits 777, then maybe in a single or a few attempts, you might get a perfect outcome that you have desired with a lazy prompt.

I am quite surprised to see a large number of people that I meet in real life actually handle their ‘AI’ in this manner. And they often blame, “Oh yeah, ChatGPT is horrible compared to Gemini or Claude.” I do agree that the underlying LLM / AI Provider does make a difference in terms of the quality of response (mostly the difference comes from their context engineering and harnessing these days). However, I personally believe you would be able to get pretty much the same quality of response across different LLM models or AI Chat services if you have written your prompt well.

Secondly, now that all individuals believe they can be a superhuman with such little effort, production of ‘prototypes’ ramps up but the quality doesn’t follow suit. It makes sense to build something quick and tear it down quick if you want to test your idea and share your imagination with colleagues. But when you want this to be live and interact with stakeholders, this would likely damage you more in the end.

Anyways, this is my personal note / rant on the way ‘some’ people use AI these days. And I personally am guilty of exploiting AI in the same manner. When your brain doesn’t work and you feel lazy, it is so easy to fall into this trap of “I will type something very lazy and quick, go and figure it out” and hope that AI will return something useful (Although, I know the limit of my ask, so I brute-force with n different sessions, like Random Forest and hoping one session will generate something meaningful). But because of this, I think it is now more important to remind yourself to give more thought, intention, and planning to what you want to do with AI. Otherwise, you will become hollow, and AI will take over you.

그 놈의 AI… 작작 좀 해라…