I took part in an AI Bootcamp after completing my degree during the first COVID-19 lockdown in 2020. The Bootcamp is for those who do not have relevant knowledge or solid skill sets in Machine-learning or Deep-learning. It took me 13 days to go through the courses and work on a capstone project. It was tough, but eventually, I passed the exam and secured my first job in a Malaysian startup in the same year. My official designation was Junior Deep Learning Engineer.
Unfortunately, after one year, I decided to resign from my company in early February this year. Here are a few reasons I quit the company and the AI industry.
1. I am a software engineering graduate without much AI/ML/DL knowledge.
I majored in software engineering at my university. Back then, people used to talk about AI, Machine-learning, and Deep-learning — The competition I participated in used AI, the champion’s project was in AI, etc. I was interested and admired those competition winners and wanted to explore more in this field. It turned out it was not easy for me because I had no advanced skills in mathematics, such as statistics, calculus, algebra, etc.
Working in the AI industry means I must go through many professional scientific articles and papers to pick the library or tool for any AI project. Moreover, I hate reading someone’s findings since they consist of advanced mathematical formulae that I do not understand. The knowledge I learned from the Bootcamp was very simple/basic. It is not sufficient for me to go through each detail written on the paper.
2. An AI system must also have a stable backend system to work seamlessly.
In my point of view, backend and software engineering play a crucial role in almost all AI projects. An AI system must have a stable backend. Without it, AI will not exist.
I admit that I am not strong in software engineering, so I would take my time to learn more stuff — the basics or the foundation in the backend.
3. We need more and more resources for the AI system.
People always think AI is a piece of code or software that can find on the Internet. It is partially correct.
A. Datasets
What if you are about to work on an AI project that deals with sensitive data? Where and how can you get the data? Does the party/organisation(s) provide the data you need? How many datasets are enough? What do you do if you do not have enough data? Do you have resources to annotate the data for training and testing purposes?
I encountered the situations above when I was in a team that worked on an AI project for a bank. It does not matter what kind of algorithms or programming languages you implement in the system. The biggest enemy here is data.
Unless you have answers to those questions, otherwise there is a high chance for you to fail the project.
B. Time
Time is our greatest enemy too. We know that training a model might take a longer time to finish. Sometimes it even takes longer than our expectations. If something goes wrong halfway, you either retrain the model from the beginning or continue training the model only if you are lucky. Even worse, the accuracy is lower than your forecast even though you have made the effort for so long. Is it worth it? You will always be facing these two situations: underfitting and overfitting.
Could we use transfer learning to save time? Well, it depends on the project. Not every pre-trained model suits the case.
C. Other resources
Training an AI model might need a powerful computer. It even squeezes your resources. If you or the company you are working with does not have a powerful GPU server, a vector machine, or any infrastructure for you to train the model, the mission is a failure.
An engineer is doing too much work most of the time. With lesser resources, an AI engineer might even take multiple roles in a project — they might cover the data engineering works such as doing ETL and other out-of-scope works. If you are not familiar with the job scopes of a data analyst, data engineer, data scientist, and machine learning practitioner, see more at this link.
If you are the only one who works in a team: annotating the datasets, creating a neural network, going through all the possible papers or articles, fine-tuning the neural network, working on transfer learning, training, testing, etc. Please rethink or reconsider what you are doing. If you are making money and you enjoy it, go ahead. Neither one? Ciao.
4. Lack of guidance.
Sadly, I had no senior to guide me throughout my first career.
So, I did not know what exactly I was doing at first, but slowly I did some research and sought help from teammates, then I had my direction.
Another biggest enemy is directionless. If you are a disciplined person, you certainly can have your direction; if you are not, it would be difficult.
5. Company’s misdirection.
If you ever experienced your company kept delaying your salary or refusing to pay your compensation. It is most probably there is no new business coming into your company.
Company mismanagement is also one of the lethal reasons you shall rethink your career future. It also caused me to lose interest in AI. When you feel no growth in your career or are facing a project/salary delay, maybe you shall stop and think carefully.
6. My country is not ready for industry 4.0, especially in AI.
I have always seen people talking about industry 4.0 and how it could help our nation. Well, it is half correct and half bragging. People tend to believe the hype, but when they participate in any of it, they will feel pain. Furthermore, not all companies are ready to adopt AI technology, here leads to the supply and demand problem. The lesser the demand, the lesser the supply. Therefore, the lesser the job opportunity.
If you are familiar with South-East Asia, you might think Malaysians can work in Singapore to explore AI technology. It is sad because the AI market in Singapore has also become saturated. Talents barely could get an AI job unless they were an experience AI practitioner.
Malaysia is one of the largest export countries in South-East Asia, we export talents. You heard it right! We export 500k talents each year. How could we build an AI nation if none of them were left in our country?
Other countries have successfully implemented and deployed 5G. We are still arguing about which company should tender the 5G contract. It is just hilarious.
7. Blockchain technology is rising. I can experience more from it.
I could experience and feel the result directly from blockchain technology, especially in cryptocurrency. I bought a coin and could still observe its value. Unlike AI, it is a very scientific thing. To a certain extent, it takes up my time and passion. It is just tedious. Blockchain or crypto is so much fun compared to AI. It is most probably we are dealing with money. Everyone likes money and pays more attention if they invest in such high-risk investments.
Also, we do not need a powerful GPU to code and deploy the smart contract (it requires when you want to be the node/validator). We can write our contracts and test them on the test net. It is much simpler than AI. Even a non-technical person can do this.
That is why I am switching to the blockchain field from AI because I see its values and the great potential in my country.
Summary
It is hard for developers to see a better result if the organisation or the party does not provide sufficient resources to build the AI system. Based on my experience, it took time to go through the processes. It does not work for me because it does not match reality.
As a junior deep learning engineer once, I spent more time learning, trying, repeating, and failing. The results always came below the expectation. I can say that it was not a good experience anyway. I was just completely brain-dead.
Please do not get me wrong: I do not encourage people to switch to other fields from AI just because of this blog. This blog is not technical. The reasons why I quit the AI industry are so obvious. If you are suffering or have been in any situations mentioned above, you should stop and reconsider your future.
Unless you are lucky to work in any company that provides various resources for AI projects like Meta or Amazon, you can stay and experience the processes of solving a business problem using AI.
Will I return to the AI industry? Nope, at least not at this moment. I am not saying I will give up AI. I will still catch up with the latest technology anyway. It is good to know about AI. We can do a small project as a hobby. Well, I wish I have time to do this.
As fresh graduates, I believe most of us have struggled in our careers. We may want to try something new or something that sounds cool. However, things do not work our way. Bear in mind, when people introduce something that sounds awesome, please beware of it — because it might just be a waste of time, especially things that came from recruiters. If you are not lucky, you keep suffering; if you are fortunate enough, you can switch a job.
Switching to another field also requires serious consideration. I have gone through all this, and it was tough. Sometimes hesitation will cause decidophobia. All you need to do is “Don’t think, just do it”.
Remember this: AI is not something you can do alone. You will need teammates, expert reviews, resources, infrastructures, and passions. Do you have none of them? CIAO.
Please share and like this post. Don’t forget to follow me. I would write more stories not just in tech but also life.
Thank you!
