As the title suggests, it is indeed a very small matter, but it has sparked some thoughts in me, so it is still necessary to record it.
The origin of the matter comes from the recently (it should have been for a while) highly regarded programming tool that combines AI, #Cursor. I didn't jump on the bandwagon immediately, mainly due to being busy (or lazy) lately. A couple of days ago, I finally found some time, downloaded it, and since I wanted to write a small tool recently, I played around with it.
Let me explain the content of the tool I plan to write:
Using Golang, especially its multithreading features, I aim to quickly summarize specific areas of content from a large number (more than 10) of Excel files into a separate Excel file in a predetermined format.
The original idea behind this is that I have a little bit of 🤏Golang foundation (I can write a few Hello Worlds), and the convenient multithreading capabilities of Golang, along with the fact that I happen to need this tool recently. (In fact, I have already implemented it in Excel using VBA, but it always takes a lot of time to summarize, and the machine basically freezes during the process).
After installing Cursor, what came into view was an editor almost identical to VS Code, which left me a bit at a loss. I first described my direct needs to Claude, which generated a framework that I then pasted directly into Cursor. To my surprise, while I was typing new code, Cursor was able to provide me with suggestions that almost matched my needs based on the context. With just a few tabs, I could write code blocks that would normally take me half an hour to an hour. Especially for some more general and commonly used blocks, like makefile, I created an empty makefile and then directly let Cursor generate the content for it. Amazingly, I hardly made any modifications, and it was ready to use. I initially just wanted to dabble a bit, but unexpectedly, once I started using it, I couldn't stop and ended up spending over 4 hours making my tool usable.
The above explains my attempt with Cursor. Now, let’s discuss the main content of this small matter.
I have a tech-savvy friend (an architect at a software service provider in a certain industry) with whom I often chat about related technical topics and industry issues; we generally get along well. One day after I tried Cursor, I happened to run into him while having coffee, and I excitedly talked to him about this tool. Before I could elaborate too much, he interrupted my enthusiastic monologue and poured out his opposing viewpoints.
- "The premise of these tools is that you have to upload almost all your code, which is very unsafe and could even involve legal issues."
- "If used internally within a company, it will definitely involve copyright issues, which could lead to legal disputes."
- "For example, when you install software, if you don't restrict privacy, you will naturally feel that the other party knows you very well and will suggest what you want to do next. Isn't that similar to PDD and others?"
- "These things can only serve as references; they are fundamentally untrustworthy and not as smart as you think. Their output completely depends on the data they are fed, and since humans make mistakes, machines will definitely have issues too."
- "Look, our country has already banned these AIs, which is clear evidence that the government has recognized the problems involved."
- "This is even scarier than bypassing the Great Firewall!"
The above quotes may differ slightly from his original words; I might have missed or confused some terms, but the general meaning should align. His unyielding outpouring left me quite astonished, particularly his resistance to AI. I believe he may not have used generative AI at all, or very rarely; and he probably hasn't even touched tools like Cursor, yet he made these sweeping judgments.
In my view, a good technical expert should have an open-minded attitude towards new technologies and should actively try them out. Of course, any new technology will have its advantages and disadvantages, or shortcomings, but one should not completely deny it because of some flaws and issues. A conversation of less than ten minutes changed my long-standing reliance on his opinions. I couldn't persuade him, or rather, I didn't want to; I don't know when it started, but I have become unwilling to convince others. I thought of some other topics and hastily ended the discussion on this topic; I probably won't bring up discussions about such tools in front of him again.