As part of our ongoing AI In your Community series, I sat down with Dr. Qiang Yang, Astrophysicist and Computer Scientist, Chair Professor in the Computer Science and Engineering Department at Hong Kong University of Science and Technology, to learn about his work in machine learning and to get his advice on how to enter the field of AI.
Tara Chklovski: What research areas are you working on?
Qiang Yang: I’m working on developing algorithms for machine learning. Machine learning is where you train a computer to do things that you do and are an expert at. The way we teach computers is by first generating lots of examples — this is known as big data. And this data is usually very special. It has annotations that are provided by experts such as doctors, teachers, and so on. But these annotations are very expensive and tedious to get. My research is figuring out how to alleviate the involvement of the expert to make machines learn things more automatically.
For example, if you teach the computer that this is an apple and that is a pear, it can infer that because pears and apples are fruits, they must have similar properties such as being delicious! So, machines will have the inferencing ability and will be able to adapt a model such as a prediction or classification model — from one domain to the next. This ability of transferring knowledge from one domain to a new one is known as transfer learning.
TC: Thanks for giving that overview of ML and what problems you are tackling! What advice would you give children as they try to find a problem in their community?
QY: I think children could think about how they can teach computers! They could think of the computer as their younger brother or sister and teach it to do the things they can do. They would have to learn a simple programming language and then they would need to learn about algorithms. They could then use some simple, learning algorithm to teach computers to do things like learning from pictures or text or voice or maybe even learning to read! Even though it could be a very simple task that the computer does, I am sure they will get a lot of delight from having created it themselves.
TC: That’s very interesting! So how would you define a good problem?
QY: A good problem has a few key attributes. First, the problem has to be important, impacting a lot of people and benefiting society. Secondly, it must be challenging; it should not have been tried before, or it’s not obvious to you or your parents how it can be solved. For example, how would you teach a computer to recognize the voice of a chicken?! Some chickens are happy, some are frightened if an eagle is coming, some are ready to lay eggs, some are sick. How do you tell the difference? It’s not obvious, right? it’s not obvious even to your parents or people around you. The third attribute is that it must be decomposable – you should be able to segment it into little milestones and steps. For example, for the chicken-voice problem, you can first record the data – the chickens’ voices. You can then observe what they do. So for instance you can label and annotate the data when they lay an egg. You can use a voice‑recognition software to convert the voice into a digital format. And then you can pass that on to a machine-learning software to simulate. It’s step by step; it’s decomposable. Fourth, you must have access to data. And finally your problem should be easy to describe to people. If it’s very complicated to describe then it may not be an important problem!
TC: The last attribute is very interesting! So what advice would you give children about designing a good product?
QY: So, first they have to be skillful at using tools. So, for example if they want to create a wooden shelf, they should know how to use a jigsaw, sandpaper etc. They may have to watch a YouTube video or take a woodworking lesson to gain the basic skills. If they want to develop something on the computer then they will need to learn how to program. Program a little every day, and keep your skills fresh!
The second thing is to be curious. The most curious animal is a cat. Be like a cat or like the tiger. Always wonder how things work and don’t work, and if they can work some other way. Don’t think like an individual in society, but step outside of society. Pretend to be an alien coming to earth and ask: why do you do things this way? Why not that way? It’s like teaching a computer!
And lastly, be confident. Believe that you are different from others. Believe that your difference can make a difference. A lot of people are not confident and even though they are curious and have the tools — they don’t dare to try it. So be brave and know that because you are different you can make a difference.
TC: What do you find fulfilling?
QY: I get satisfaction from knowing I can impact many people’s lives. For example, we are working with the Chinese social media company Wechat that has 1 billion actual users who use it every day! Wechat has a social impact group that launched the Voice Donor project two years ago to help blind people listen to a novel. Currently there are only a limited number of audiobooks and it takes a huge effort and lot of resources to create one. WeChat used crowdsourcing through which users could record their reading of a piece of a novel. WeChat collected 2.3 million audio clips, chose the best readers, and pieced together whole books read at very high quality for the blind. They then took this to hundreds of blind students and the students were very moved. When I saw that video, I was very satisfied.
TC: What do you find difficult?
QY: I find it difficult to mobilize people. It takes extraordinary spirit and effort to achieve something like Steve Jobs or Bill Gates achieved. They saw something that other people didn’t and believed in it. So that’s why I say pretend to be an alien because I find a lot of people around me are ordinary people, ordinary humans. I try not to be a human!
Another thing I find difficult is to assemble a good team. Often people have a good idea, but they don’t have a good team. So they can’t put things down on the ground and their ideas just disappear.That happens a lot in history, and it has happened a lot to me too.
TC: So how do you overcome that difficulty?
QY: Just by keeping healthy, living long, and hoping for the next chance!
TC: How do you think AI will strengthen societies and communities?
QY: AI is a very powerful thing, and it will do more good than bad for society. For instance many of the activities we humans do, are not enjoyable. Children will give you a long list if you ask them about what they don’t like. If you ask an adult on Monday morning, they will tell you that they don’t want to go to work because it is boring; they don’t want to wade through traffic; they don’t want to do things that are not productive and do not help them realize their potential. And they don’t want to do things that machines can do. For example, nobody takes their clothes to the river to wash anymore because we have machines. So AI is there to liberate us from the inhumane work.
AI will also help with connecting people together, keeping them healthy and happy. For instance by giving us lots of good movies to watch! Maybe in the future AI’s will make more movies like those of Steven Spielberg by copying his techniques and reproducing them by ten thousand. And then we can have more good movies!
TC: So what specific field of AI excites you the most?
QY: Computer games are very fun, but AI is the most sophisticated computer game! If you’re working in AI you get paid to play a game every day. That’s the best job you can get!
TC: That’s an awesome plug for AI! So, what do you think is the best way for children to learn more about AI?
QY: I think children should know that to learn something is enjoyable, but to teach somebody is even more enjoyable. That’s why when you ask children what they want to be, some say policemen, but many say they want to be a teacher. They believe a teacher is grand and wonderful. I think teaching is the essence of human happiness. You extend your own life when you teach somebody because your idea is living in somebody else’s body. I believe that even children realize this unconsciously. So, it’s very exciting to teach computers what you know and make it smarter than you are.
TC: Is that why you chose AI?
QY: I grew up in Beijing, China and and studied astrophysics at Peking University. I studied cosmology and solar physics as an undergraduate. And then I came to the US to continue in this area and get a PhD at the University of Maryland. But then I discovered computers! That was when computers were still using cards and paper tapes, and I found that fascinating. So I switched and got my PhD in AI. I worked for many years after that in Canada, and then I saw Asia was awakening — especially China. Many students wanted to learn and so I chose to come to Hong Kong – a place like the US and Canada but close to China! So that’s a brief history of my life!
TC: I just finished reading the Three-Body Problem. What do you think about it?
QY: I’m very close to that field. I think it’s not entirely impossible for that to happen — possibly in the distant future.This book really opens a window into the future. It’s a combination of physics, human intelligence, intelligent beings, and communication. Also, it has two hierarchies in contrast to other science fiction works. One hierarchy that has been explored in the past is the hierarchy of dimensions in physics, like gravitational forces and the general relativity field. That’s the physical world. But there’s also a hierarchy in the book that has not been explored before — intellectual hierarchy. It’s likely that aliens are living on a higher intellectual hierarchy and when they come to earth, its likely the interactions with us could be similar to those portrayed in the book!
TC: And on that grand note, we will close this interview. Thank you so much Qiang for sharing your golden insights, philosophies with us!