4 minute read

Ok, ChatGPT is flooding the media, people are crazy about it. As an guy who codes for food, why not try it then.

How to start

Just google `open AI chatGPT” and you will be guide to the chatGPT link, unfortunately, at the moment, it is currently geographically blocking Vietnam user, there are several workaround you can find out there.

Let’s get started

The feel

Obviously, as an machine under the hood, chating with it is not something like the human, however, for the purpose I am wanna try, it is interesting than using google search, but it is lke using the google assistance, except that this bot is not internet connected, but use their historical trained data. I guess openAI have provided it massive knowledge.

One think I was always wanna do but hate to learn is the GUI, like Linus Tovard spoke in TED session with Chris Anderson. Then, I have asked the bot to generate the code of simple GUI program (adder of 2 input numbers). As first, to be frank, I am scared of what it can do, it provide the initial version, then I complain, and it provides the revised version, and it did excellence for several round. The fololwing is the simplified conversation between myself and the bot.

Well it is really terifying, but in the end, it is just the massive if-else, isn’t it

I am a fool, I am always think of the neural-network works like the massive if-else mesh. After couple of hours playing with in, I am convinced about the capabliity of the context in generated

  • My goal is to make an salary calculator in python with GUI:

    • I wanna learn a bit of the GUI in python so I asked the bot to generate the simple GUI app, with some input textbox and one calculate button. I copied the python code it gen and run it run pretty well.

    • I see the font is quite small, so I asked chatbot to increase the font size. Well It remembers the previous code and it generates the code to change the font size. Moreover, it also explains to me the code like a teacher, cool huh? I like it.

    • Tweaking my questions some more rounds, and it performs like an expert in python. The content it generates having the consideration of the whole discussion, not just the dry ass Q&A.

    • However, when it comes to some bugy code it generates, I copied the error to the chat box to challenge our bot. It started to behaves like a junior, haha. I have tried to simplify my question, it seems to help a bit, but it cannot hide the fact that the code it generate is not from itself but the mix of some unknow expert in stackoverflow (haha).

  • Actually, my plan of learning haskell is abit behind the schedule, because I am so stupid to understand the monoid, monad, May be,… So I would like to check if it can be my mentor during the learning. So was asking the following question.

    • In the ELI5, what is the functor in category theory. The following image is the outcome. Impressive huh?

What we should prepare as an engineer?

For me, with above 2 simples example, the chat bot like GPT really awesome to the point that would put some pressure on us, chat GPT seems to to compete with us, the engineer in these tedious job (like implement some low complex requirements). That may not immediately affect our job now, but the junior engineer would be in the difficult market now. Frankly, if I was able to use my 2 hours of engineering to tell the GPT to make some simple app, which may take a whole week or more of a junior, I would prefer worknig with chat GPT (I hate working with human haha).

Personal verdict?

Think of how magically our brains are, I don’t think we could build the “skynet” that is superior the humankind intelligence in total (and the skynet without weapon connected is harmless, isn’t it?), I think the main reason the chatGPT is disconnected from the internet is to ensure that it cannot learn the negative stuff from the humankind, thus their purpose is achieve with just a little side effect. However, as an egineer. What I can see that it can generate the simple code that it would take me a day or so to do, in just a few minute. Thas is really impressive. Then it is that, the chatGPT cannot handle the much of the complexity like a middle engineer, but for simple task like writing some small module we normally see the entry level engineer did, I am afraid that these job will bee soon replace if the cost/performance of chatGPT is considered. The it is not the end of the our career for sure, for instance, before the calculator is invented, we still doing the math on paper. And when the calculator and computational science is rolling out, it is not mean that we can fire the mathenmatician or we can skip the Algebra 101 class in the college. ChatGPT or equivalent tools will be the powerfull tool to do the tedious job of the programming, like the calculation remove our stress of incorrect calculation, or like the computer spreadsheet (and database recently) replace the paper-and-pen book keeping.