People walking down the street take on different roles depending on my level of interest and my own situation. Even when they are simply going to work or leisurely heading to a department store, they can become enviable people with clear destinations, using the road literally only as a passage, when I have no destination of my own. When I am on my way to deal with something unpleasant, they can look like people enjoying peaceful everyday lives. And when I have taken a day off, or when I am not going to work and have come out for a walk, they can become busy people with no time, hurrying toward somewhere.
Doing something with AI is similar. The mere fact that one is using AI, and the time spent using it, places one on the road. You can squeeze AI’s abilities with the will to create something specific, or you can wander with only the vague concern that everyone else is using it, so you should at least know how to use it too. They say that if you put a ballpoint pen in someone’s hand, they want to scribble words or drawings somewhere, and if you put a hammer in someone’s hand, they want to hit something. In my case, I attached sponge to a hammer and use it to knock used coffee grounds out of my coffee machine in one piece. But that is because I did not have a rubber mallet and intentionally kept it in the kitchen, not because there was a hammer in the kitchen and I thought I should hit something with it.
AI is the same. Unless you put it somewhere visible with a purpose, it ends up scribbling or damaging furniture. And unfortunately, AI as an app inside a phone is always visible, so it easily becomes a victim of the thought, “Let’s try doing something, anything.”
We live in a world where publishing with AI has become a problem. The problem is not books made using AI. The problem is using AI to make something pretend to be something else. In my view, the people causing those problems have at least seen through AI as a tool. They used AI simply as a money-making device, and exploiting a loophole is, if not quite fraud, close to deception. In that sense, they used AI for an immoral way of making money. Still, the quantitative supply that made even that deception profitable would have been impossible without AI, so I do not think it is too much to say that they saw through it.
Through this project, I came to understand that thing called purpose quite painfully. Just because something is hard to use does not mean it is impossible to have purpose. And in fact, it is not even hard. I thought that showing that might help other people approach AI more easily, at least in programming. From there, I also wondered whether, instead of merely boasting that the firmware might be usable, I could show that an actually usable product had been made by filming myself writing inside that firmware.
The conclusion was: <Do I really need to…?>
I wrote a couple of Brunch posts on the ZeroWriter using my firmware, filmed the entire process from beginning to end, and uploaded the videos to YouTube (YouTube link). But since the hardware was not mine, it did not seem to have much impact. After showing it two or three times, I did not feel much reason to keep filming videos where only the black, round, square-looking letters inside it changed and everything else looked the same. So I thought those two videos were enough. I think simply showing that this much is possible became a good actual-use review for both the new firmware and AI.
At the moment, I use only ChatGPT and Codex. Gemini encouraged me to begin, but somehow the two that remain are those two. I am working on two new projects at the same time, and even now the work is divided in the same way: ChatGPT unpacks and organizes complex concepts, while Codex handles coding and calculation. I now make the apps and computer programs I need and use them myself. AI is now an employee that does assigned work. It is no longer a friend. I checked this several times, but in the end, it is a temporary relationship that ends as soon as the chat room changes.
It is hard to believe that all of this happened in three months. From mid-to-late May to mid-August. During that time, I made Korean firmware, then updated it into a version that can use every keyboard in the world. I added Wi-Fi synchronization and automatic updates. For people who might feel uneasy about installing it, I made a web browser version so they could try it first.
And the AI work continues.
When there is no purpose, it is fine to rest for a while.
Because it is a tool.
Original Korean article: 실사용기는 실사용이 최고