5 Things About This Female AI Engineer You Never Knew


Tech Storm had the pleasure of having an exclusive one-on-one interview with Annabelle Kwok, the founder of 2 Artificial Intelligence (AI) startups - SmartCow and NeuralBay. She has been featured on international media such as Yahoo! News, NVIDIA, Tech in Asia and has also spoken at huge regional conferences such as Echelon Summit, Deep Learning Summit and Innovfest Unbound.

TS: What inspired you to start NeuralBay?

AK: It is quite a long history…when I was 19 I spent two months in Africa. I did not have many plans there, but then I saw the kids with fat bellies running around; and when I went down to the villages, I saw that the women really zuo bo (idle). They are already very poor, so it is not so much about women empowerment, you can do this, but it is more of: you are already so poor, so it is better that you work, use your extra set of hands. So I started working with non-profit organisations and local universities there to try to alleviate all these kind of problems - whatever I can help I will do my best.

Within my team of about 20 people, about 4 or 5 of them had malaria, and everyday I was eating doxycycline – malaria in Africa is rampant – as well as using mosquito repellents. But when I was using the mosquito repellents the locals there asked me “what’s that?” And I was quite puzzled that they did not know about mosquito repellents which has been around for a long time, and also quite cheap. And finally some of them stated that they knew about the mosquito repellents but could not afford to buy, so the companies do not sell these repellents, and those who try to sell the repellents for cheap will close down after a few years because they could not do this heavily subsidised model.

So a lot of them have malaria; at first at my table, I did not know that [many had malaria]; one of them was lying on the table like that (placed face on arm). I thought he were sleeping so I said “Can someone wake him up? Why is he sleeping?” Then others said “He didn’t sleep, he just fainted”. After we woke him up I realised that he had malaria and told him to go home and rest. It is like the same tonality that you tell others when they have a fever or a cough – “Oh you have a cough/fever? Go home and rest!"

So that was the ta-toouch moment for me - it’s like, why is it that in Singapore or elsewhere, we are pushing the frontiers of science, of medical research, when the people who really need it the most, the people who are fainting and dying from diseases, don’t have access to it? That was my epiphany, so I really wanted AI to be easy, afforadable and accessible for SMEs (Small and Medium Enterprises) to use, because if you look at those who do AI right now it is all the big tech companies – Amazon and Google – and there is nothing wrong with that. But then you are going to see a big technical disparity in future and this technical disparity will lead to financial disparity and all these SMEs are going to be overwhelmed. So how do you actually help them stay relevant? So I believe that you don’t always have to give back in terms of money and time, you can also give back in terms of knowledge and skill. I know that I am good at building stuff, so the reason I started my own company and not work in a corporate job is because I really want to build stuff and have a strong say as to where the stuff I build goes to because if I join a corporate company the stuff that I build will go with them. So it is my feeble effort to change the world. <chuckles> I wouldn’t say change the world per se but at least help the SMEs who lack the resources to do it. Why SMEs then? It’s because you see most people working there are not the extremely rich - they are lower middle to upper middle income people and a small difference in terms of how much they earn, in terms of the bonus that is paid, if you pay them a bit more they will feel the difference - better food on the table for the kids etc. - so this is how I am trying to help.

TS: How does NeuralBay help companies?

AK: So we have two arms for NeuralBay - the first arm is consulting, because we have to be a business, we have to sustain. We do a lot of projects - we have one with Ferrero, and we do many other big projects as well - those are our fat pay checks that sustains us. But also we have the marketplace where we try to modify things to make it more generic for the mainstream SMEs. So it’s more kind of, when these SMEs want to search for something such as drawing virtual parameters and counting the number of authorised and unauthorised people in a zone, we can just "click click click" and get it done. You see, you don’t need to know hardware engineering to use the laptop, you don’t need to know software programming to use a mobile application, so when you want to use AI, why must a small business hire their own AI engineer and coder when they can just buy something off the shelf and "click click click" and get it done. So this is something I want to build for them. But how do I actually juggle both of these two arms (they are two very different things)? So let me first explain to you bit of AI, it’s like if you have a neural network model to detect body posture, it can be used in security industry to detect a person crouching in the corner looking nervous and suspicious - which you can then go and investigate, and then there is another function, to use in the medical industry such as physiotherapy and rehabilitation; for example to monitor a person doing his exercises correctly when the doctor is not there to supervise - this can be a more efficient use of time. You can see the medical and security industry have very different error and risk tolerance levels, so when we do consultancy, when we do marketplace, the core of what we do is learning to detect body posture or faces or gender or clothing or objects like chairs. So for us it is quite okay to do both at the same time. For marketplaces, we will do one industry by one industry to make sure that we really cater to their needs.

TS: How did you get interested in AI?

AK: Long story…I used to build robots. I freelanced a bit in robot building projects, and also in university I studied image processing and neural networks. So back then AI was not the hype, it was like just another boring module, but I did a lot of it. I even sat in on a PhD module during my undergraduate because there was not enough students to make the class so the professor agreed that to set easier test papers and I was “ okay lah!” So I had the background foundation already, and for one of my projects we had to masking-tape a laptop to a moving robot that is chasing a guy in a room. There's a camera on the robot and when the camera detects the guy, the robot will fire a paintball at him, but the guy is running, so you cannot just fire the gun because the guy would have run away by the time the paintball reaches him, so there is already a bit of prediction analysis, bit of image processing, and that is my background. But to me, it wasn’t like wow, that was AI, it was more of just another cool feature that I could to add to this machine. Back then, there was actually no device such as heavy duty kind of processing to run AI algorithms because laptops were build for Microsoft Word or a bit of Dota gaming. So that is how I started my first company (SmartCow), the AI company that actually build hardware devices for people to deploy. Imagine you want to do image processing, you want to recognise someone, if you have six-figure money sum you can connect to your server in your server room but if you don’t have that six-figure sum then what you are going to use? Laptop with masking tape? Hence for my first startup, we develop a small board that is as cheap as a laptop, if not cheaper, smaller, lighter; something that you can actually use. However, after that, I realised that hardware is a very tough game, it is a big boys game, it is really economies of scale. Moreover, I also have a lot of software background - I graduated in Mathematics. I recalled when I wanted to sell my boards in the past, others presumed that they would get facial recognition, body posture, gender, clothing features as well, but I told them that that’s not the case - it’s just like when you buy a laptop it doesn’t come with all the games, you have to go and buy from the game providers. And back then for these companies, each one only focused on one thing, so a facial recognition company would most likely not do speech analysis. Thus you have to find someone to integrate all the different kinds of things, and this is very hard because all the “integraters” are usually the old folks. And therefore most do not want to buy and these people are from international MNCs. So I decided that I could fill this gap and so NeuralBay is software based.

TS: Do you think that AI progressed a lot in the past decade?

AK: AI is over-hyped, seriously over-hyped. In terms of media hype it has boomed a lot in the past 2 years. In terms of technological development, we are still at that debate: if you look at the number of research papers published, yes, there are a lot of papers, but in terms of what is actually concrete and can be implemented, we are still arguing about it. When you come out with an idea or you make a new discovery, people will try to challenge you, and you will challenge back. So we are still at that phase that when you write a paper, someone else will write a paper to slam your paper.

TS: What role do you think AI will play in future?

AK: You see, humans still need to eat, sleep, and sh**. You only have 24 hours a day, no more, no less. No technology will help you increase or decrease the time you have. It is more about how you actually go about doing the things that you do more efficiently. So if you think about it, in the past humans walked from farm A to farm B, and as time progressed we use the bicycle, then the car; maybe next time we can sit on flying drones, who knows? Basically, I don’t think AI will be a big game changer or a big disruptor - people will still go about doing the same tasks, it’s just that there will be minute tweaks and changes in terms of the efficiency.

TS: Do you think AI will potentially threaten humans?

AK: Here’s the thing: I think that many big names are saying news [that AI is a threat] to quash the hype around AI, because when people don’t understand things, it is human nature to over-think and react - and I will draw this very bad analogy: imagine you come into a country and you see the Red Indians, you will go ahead and kill them. So when we see AI potentially taking over our jobs, we will be very fearful. Some people might think that AI is this almighty thing that will take over the world; and that is the main stance that the media is taking right now. But this is over-inflated. The main thing is that the Red Indians are just peaceful people and don’t pose any threat, they don’t deserve to be killed. So it is a matter of people not understanding and over-thinking the whole thing. Therefore, I think these big names are just trying to quash the hype - because media journalists are not technically trained, whatever they can sensationalise, they will sensationalise. So as long as the car move up 1cm, they will say that we have a flying car! And that’s the problem - in the past there were two times when we had an AI winter - a hype and rise of AI which then crashes. The AI hype has stayed silent for a while and now it is back. And when there is a big hype about AI, all the investors will be willing to give out money, and the researchers would want the money can fund their research. The researchers will start taking the money, but in order to do so, they have to promise the investors certain deliverables, and they start overpromising. Eventually they realise they cannot deliver on the promises, and will start pulling out - it is just like the financial crisis. So we have to be very careful not to jump into another AI winter again, and that is why I believe so many people are trying to quash the hype.

TS: Do you have any advice for aspiring female technopreneurs?

AK: Don’t do it for the glory, don’t do it because it is cool, just do it because you have a genuine interest in whatever you are pursuing because bubbles and hypes comes and go. Right now it is AI and blockchain, but it will come and pass, just like last time when the trend was IoT. So don’t chase the bubble, if you like chicken rice, stick to chicken rice, just because salted egg fish skin is the in thing now doesn’t mean you have to keep eating it. Go for something that you like.

TS: What are 5 things that most people don’t know about you?

AK: I went to North Korea to run a marathon two months ago. I got to high-five a lot of North Koreans along the way, they even give me a kiss on the cheek! I can also juggle and spin plates as I used to be part of a circus. Third thing…the US Embassy flew me down to meet President Obama. So I have videos of Obama’s sexy bottom <chuckles> on my phone. Reuters News caught me taking the picture and I appeared on their headline photo. Oh and another famous photo that was always circulated was the one when the US embassy was live streaming and they were telling me “Annabelle - smile more smile more” so I was like <gives the huge grinning face>. Fourth one would be...I play a lot of music - I play the guitar, harmonica, piano, saxophone, djembe, and violin. Lastly, I waste a lot of my life away on Youtube! So when I tell everyone that I am very busy, I am actually wasting my time on Youtube. Shhhhhh!

By Jesse Tan/9 July 2018 14:00 Singapore