We saw the rise of Large-Language Model (LLM) AI a few years ago. And it is a controversial topic about the morality use, privacy, plagiarism, economy and so on.

In today’s article I want to share how the AI will affect our lives based on what I read or heard on the topic. You may agree or disagree, it is just one view on the topic and not necessarily scientific facts. I am just a beginner on AI world.

Then I will share my experience on using LLM for code generation and how I view code generation in general.

AI will replace all our jobs, maybe not

In the age of AI, many are wondering “is AI taking over our jobs?”

I hear a lot about AI is going to take our jobs. Because of that some started to speculate about how the future will look like. I heard about the universal basic income, and that means when robots and AI do most of the things we humans do now, we will get a lot of people with no useful skill. So instead of letting them be, we just give them the “universal basic income”, so they keep going. People describe this future with different details, but the main idea is that “universal basic income” will be given to people who lost their jobs because of AI or can’t get a job because of AI.

Another and more optimistic view is looking at what happened in history during industrial revolution, internet and even smart phones! Although back then it was argued that those technologies will take a lot of jobs but what happened in reality it created more jobs than before. And smart phones cameras did not kill professional photographers, we have more of them now.

And the debate will continue for a long time. Only time will tell who was right. However, I am sure AI will change our lives a lot. And I can see AI can do a lot of the stuff we used to do like coding, generate photos and even writing an article like this one. Maybe I wrote this using AI. (BTW, I used zero AI in this article)

My experience with AI tools was not good. And I am talking about AI for consumers like chatGBT and Gemeni (below I will just use chatGBT to include chatGBT, Gemeni and copilot).

AI and Coding

I use chatGBT to do some code generation, I know there are better tools, but I did not have a chance to explore them.

I am in the middle of master’s degree of “IoT and Embedded systems”. I have some coding assignments, and I get some help from chatGBT (if you are an educator reading this, I am sorry to tell you this, all students are doing this including yours). A lot of time chatGBT generate wrong codes but I can say 80% to 90% correct and sometimes 100% correct.

If I am using a popular platform like Arduino or raspberry pi, chatGBT can produce good codes. However, when I tried chatGBT for MBed OS (it is a type of RTOS from ARM), I did not get a single line of code correctly. You know why? I think the reason is that not a lot of people are playing with MBed OS and posting their work in the internet. And chatGBT shows you the source of its prompt and the only website shown as a source of GBT prompt is the MBed OS official website! So how chatGBT can produce code that it knows so little about?

A few weeks ago, I attended a one-day seminar “IoT and digital twin”. The instructor showed us how we can code raspberry pi using chatGBT, it was great, and the instructor even said, “I feel sorry for the time spent on computer science degree”. I told him about the issue of MBed OS, then he said that I might not have asked chatGBT the correct questions and continued “there are people out there who are Prompt Engineer who knows how to ask chatGBT the correct questions in order to get the results you want”.

The point I am trying to make here is that, as long as you are working on a very popular platform where lots of people post their work on internet (like Arduino an Raspberry PI), chatGBT can search the internet and generate the code for you. On the other hand, if we are working on advanced RTOS like MBed OS, Zephyr OS or ThreadX, chatGBT can’t help you much.

My experience is limited at this time and mostly on embedded systems, I am sure there are other examples where chatGBT can work on hobbyist level but not the advance level.

Conclusion

If you make it this far, I am sure you can see that I am still not convinced of chatGBT to be super helpful, not in 2024 at least. Who knows, maybe I will look at this article in 2035 and laugh.

In today’s article I only focused AI and coding as an example. The debate of weather AI can replace other skills exits as well. examples are learning languages, writing, video editing, song writing and so on. But I will leave the discussion of these for future articles. As this one is long enough.