Everday AI is being talked about, majority of us have been using AI daily one way or another now. However, how many actually know the mechanism of AI ?
In a layman perspective, look at an example of asking AI to predict whether a person will step forward left or right foot first to start walking. It is like a 50-50 luck guess. However, if you have a collection of data such as how many male or female uses which foot first, how many right-footed or left-footed uses which foot first, etc to train the AI model, the AI will eventually be able to predict with a higher probability other than a 50-50 luck guess. The more data you train the AI model, the higher the probability will be the correct prediction. AI is all about predicting based on probability and not a certain. Unfortunately, in real world, thing is not that binary as in left or right foot, yes or no, true or false. To cater for the multi-dimension, the AI model will need to be trained on very comprehensive data in order to produce a very good and high probability of correctness.
The above is from the AI app that I developed from scratch to predict the buy and sell signal for stock market. The first series of sell signal for STI was first predicted in July 2026 before consecutive signals appearing in Aug, Sep and Oct. The latest was 6th Oct 2026.
The above is the short trade that I carried out and still in position in Aug 2026 following the AI prediction. As seen both short positions are having at least 10% profit as of 27th Oct 2026. Due to the leverage nature of the short, the actual profit is more than 300% from the margin cost being used for just 2 months plus of shorting it. STI hit an all time high of 5.828.39 on 4th Sep 2026 and as of 27th Oct, it closed at 5,399.03, a drop of 7.4% from the peak.
Now, imagine if there is such an AI app commercially avaliable (maybe there is already) with many users. When stock markets move up, following the AI prediction will please everyone as you are making money. However, now in the case of stock market selling down, a series of sell signals from AI prediction is like in a room when someone shout fire and everybody panic and rush for that same exit at the same time. It will be a carnage when sell signals start flashing all over the place. You sell it early, you still can make profit from the previous long but if you sell it late, your profit could turn into loss or your loss could even turn into a bigger loss. AI is not human being, it doesn't have thinking and emotion, it just output prediction based on the data it has been trained. This is the pros and cons of AI. When AI predicts market going up, all making money but when AI predicts market selling down, those who are late to follow will smell blood. The above app is developed for own use as fully aware of the undesirable consequences it will lead to if it is a commercial product. Thus, do not approach me regarding this app.
Now take another example in which AI is being used in services and support industry. Take the case of a cafe selling various type of coffe (espresso, cuppacino, latte, etc), if you have an AI that upon scanning the image of a customer can recommend a top 3 choices of coffee he/she should get. In order to do that, the AI model has to been trained from data that say correlated gender, age, etc of the people to what these people prefer to drink to work out a high probability prediction. This could amount to thousands or even millions of people data in order to get a high reliable probability. To make the service even more personal and advance say to regular customer, the AI upon detecting the regular customer, immediately place the order and just wait for the customer to make payment. To do that, the AI will have to capture the regular customer's data like gender, age, day of purchase, time of purchase and so on. Basically, it is training on the lifestyle of the customer in order to make that correct prediction.
This service look awesome but it comes with a price, the data collected. These data that need to train the AI model is not simple data, it is very detailed data of a person, to the extend lifestype habbit of a person. Should these data be compromised, these customers can easily be a scam victim next. That the risk of AI and nobody can guarantee these data will never be compromised one day.
Countries one by one are starting to adopting AI in one way or another. The big concern is when Government decides to use AI, the data collected from the population to train the AI are very private and confidential data. Firstly, does the Government seek consensus from the poopulation to collect and uses these sensitive data ? Secondly, if the Government "forces" the collection and usage of these data by law, anything happen who will be responsible ? You cannot assure these data will never ever been compromised. Well in July 2026, SLA has confirmed 70,000 Singaporean data such as NRIC, past/mostly outdated property addresses have been compromised.
As the world advance in technology in using AI, it cannot ignore those risks. The AI risk is already there since day 1. Weird though, developers of AI clearly know the risk but none sound out but instead full steam ahead to roll out AI. Latyely, these advanced AI players very confidentially said they could contain the AI risk. AI in today world is in an uncharted territory, in few thousand years of human history, confident in uncharted territory only result in over-confident and bubble burst is just waiting to occur only. Human nature never change for the past few thousand years.
While we can't rule out the benefits AI could bring but it also comes with the risk. Adopting AI should be carried out in a more cautious and adapetive way instead of just go all out for it. Just like a double-edged sword, use it correctly, it help you to cut but misuse or over-confident in usage, it will cut you instead.
