It was on a Thursday evening that I met Lau Jin Ming. Clearly, he had had a long day but the foodpanda delivery rider was raring to show me his latest bit of tech wizardry — an artificial intelligence chip which is able to identify paths and steer Lau's bicycle accordingly, without the use of his hands.
Take a look here at how it navigates itself around a HDB estate:
How the AI chip steers the bicycle
Regrettably, I'm not as well-versed in technology as Lau himself, and so much of the conversation about the actual coding of the AI programme flew over my head.
However, the 22-year-old, who is also a mechanical engineering student at NUS, gamely tried his best to explain how the entire contraption works.
Nestled in a pouch on top of his bicycle was a little green printed circuit board no bigger than a biscuit.
Everything starts from this chip, which sits on a circuit board. This chip hosts the AI programme — essentially the brain of the entire operations.
The programme receives a live feed from USB camera mounted on the front of the bicycle.
And it is particularly clever because it is able to recognise lines, like those that form paths and roads, from the video feed. So for instance, if the AI programme sees — from the live feed — that the path in front of the bicycle veers to the right, it sends a signal to turn right.
This signal is transmitted to an actuator on the circuit board, which in turn moves a lever connected to the front of Lau's bicycle, and — ta-da! — the bicycle is steered right.
Here, it's important to note that the AI chip only steers the bicycle left and right. It doesn't propel the bicycle forward. For that, Lau still has to rely on his own legs to pedal.
There are other bells and whistles too.
A torch was added so that the camera can capture clearer images at night, and Lau even threw in a nifty little musical feature to help him troubleshoot his prototype AI chip.
"When it's turning left, it'll emit a certain music note, then when we're turning right, it's another music note. So when I cycle, I can hear and I know whether it's functioning correctly or not. So in that case, if anything is not functioning well, then I have to go inside the chip and just code again."
This chip is the product of a long process of planning and trial and error — very much a passion project for Lau.
He first had the idea to create this self-steering AI chip about two years ago, right at the time when he first learned about artificial intelligence and computer vision at a hackathon.
At its simplest, artificial intelligence is like giving a machine an MCQ test and training it to produce the right answer, said Lau.
"It made me realise how easy it is... it's very logical. You just have to follow the steps. Step one: Get the information, step two: give it an answer, step three: test the AI."
His first prototype materialised about one year ago, and the current version is the result of countless reiterations and physical sacrifice, including a scar from a knee injury which he sustained while testing out an early iteration of the AI-assisted bicycle.
The beauty of this AI programme, according to Lau, is that it is continuously learning and it gets better and better at recognising paths and lines over time.
With such a nifty chip, you might think that it is quite handy in his work as a foodpanda rider.
But Lau said that he does not use his AI chip when he's working for one reason: According the Highway Code, cyclists have to have both hands on the bicycle handlebars when they are on the road.
Off the road and in his own free leisure time, however, Lau likes to take his AI chip out for a spin.
"Sometimes when I'm very hungry, I just eat (on my bike). Then I just let the AI do its thing."
But Lau doesn't rule out the possibility of using an AI chip for his work as a foodpanda rider in the future. The success of his self-steering chip has opened his eyes to other AI-assisted possibilities, for instance, in programming a chip that can ring a bell whenever it detects people in front.
"That's my future project!" he laughed.
"My next step is possibly to programme the chip to detect humans in a frame, and to detect if it's within the two lines, e.g. within the path. So two conditions — got human, and they're within the path — then it will ring the bell."
Loves learning new things
Lau's interests extend beyond artificial intelligence.
The student, who professes to have "a lot of robots" at home, built one which — get this — switches off his air-conditioning unit in his room and opens his windows whenever it detects him waking up.
The music aficionado also plays seven different instruments — the guitar, bass guitar, piano, drum, violin, cello and melodica — and they were all self-taught, by the way.
But above all, Lau's greatest love is cycling, which was the reason why he became a freelance foodpanda rider five years ago.
"I saw these people cycling around and I was curious. Because they could earn money while cycling, then why was I not doing that? Because I really really love cycling."
What drew him to the platform was its flexibility (he is able to book a slot to work whenever he wants).
And since riders have to book slots to work, there are always jobs within the booked timeslots.
"For example, Company A, you maybe only do two orders but you work for five hours, but you only get paid for two orders. But for foodpanda, because of the way they stagger out the slots, right, you'll most definitely get a slot for delivery, like minimum two to three per hour."
Today, as a university student, he has much less time to cycle for foodpanda, but he still does it once a week because he loves how the work has helped him to grow as a person.
Being a delivery rider taught him to be more independent and to rely on himself (and occasionally on Google Maps), and to be tactful in the face of unreasonable customers (whom he encounters about 10 per cent of the time).
"People have to understand that actually, we are not the ones who are completely at fault for late deliveries," Lau explained.
"There are other factors like weather, or the restaurant taking its time (which is the main reason, actually). Maybe you got your own bicycle problem, maybe you fell down halfway.
There are a lot of things behind the scenes lah. But even with all these, I learned to give a good face lah. Don't let all these things affect your appearance."
But perhaps one of the more lessons that Lau has gleaned from his experience as a delivery rider is this: Delivery riding is not sai kang (a sh*tty job).
According to Lau, people have a misconception that delivery riding work is sai kang and hence they think that its only suitable for the lowly-educated or those who can't find better jobs.
But this is an over-generalisation, he said.
Some folks, like himself, take on delivery work because it's an easy way to earn part-time cash and indulge in his passion for cycling.
Lau also pointed out that at one point of his food delivery stint, he was earning S$22 per hour, which is not really what one earns in a low-income job.
"Just an example," he said, while pointing to a nearby eatery.
"If you work in F&B, maybe you might about S$7 to S$8 per hour only. But foodpanda is almost triple. I think if you work correctly, if you work diligently, and also, you don't make any mistakes, I'm quite sure you can be a high income worker."
Thanks to his love of cycling, Lau won't be stopping his foodpanda delivery rider stint anytime soon, and hopefully, this means that we will be seeing some nifty AI-assisted delivery rides in our estates soon.
Top image credit: Lau Jin Ming
Follow and listen to our podcast here