• CEIBS Alumni

    Join the largest, most influential b-school network in China: over 24,000 business leaders, and growing.



Cheng Hao: Go All In On AI!

Volume 2, 2017

A venture capitalist focused on high-tech investments, Cheng Hao worked in Silicon Valley and at the Chinese internet giant Baidu before co-founding Xunlei Technology, which listed in the US on Nasdaq in 2014. He has a degree in computer science from Duke University in the US and is a CEIBS 2016 CEO alumnus.

“What is Artificial Intelligence?

Artificial intelligence, or AI, is a hot topic these days. It’s based on the principle of machine learning, in which computers can ‘think’ without being programmed, and today you can see it being used in many ways. The most common AI application is the internet search engine. There are hundreds of thousands of parameters on the internet that would be very hard for humans to sort through to determine which is negative, which is positive, and judge their relevance. How does the computer do this? It first sets a weight, then grabs the results and forms a data loop according to user feedback, before adjusting the algorithm and parameters. This is the core principle behind a search engine.

When you browse e-commerce websites such as Amazon.com and JD.com, you will notice two or three products being recommended to you at the bottom of the page. This is another example of machine learning at work. How do these websites come up with these recommendations for you? They rely on many factors, however the most important thing it will note is whether or not you click on the item, and this will influence future recommendations. The website Toutiao.com also uses this principle.

Map navigation is another example of how we use AI. I believe most of us have used Baidu maps, AMAP or the pre-set map in our cars’ navigation systems that were made by OEM manufacturers. I never use the pre-set map in my car because it is updated very slowly; it cannot go online, therefore it doesn’t use machine learning. A navigation map may recommend we travel in a straight line from place A to B, but we don’t always follow that suggestion – maybe the road is being repaired, or there is a traffic jam. A smart map which utilises machine learning will provide a better solution next time you are travelling from place A to B – without machine learning it cannot improve; this would be a ‘stupid’ map.

Which Businesses Are Investing in AI?

Robotics is a big area for AI. There are three types of robots. The first is industrial robots, though many are not really robots at all, they are more like intelligent equipment, focusing more on accuracy than intelligence. The second type is the service robot. Let’s take Starbucks for example; to me an ideal Starbucks shop is one where a customer sits down and scans a QR code from their WeChat to order and pay, and their coffee would be made and served by a robot. The coffee menu in Starbucks is standardised; therefore it wouldn’t be difficult for robots to fill the orders. Some time ago I saw a robot which could lay bricks; bricklaying is actually very difficult work, and now it can be done by a robot. There are also robots which can be deployed in an orchard to pick fruit, and sort them according to colour and size. The third type of robot is for consumers. There are only three kinds currently available: educational robots, aerial drones, and robots that can clean your floor. The emerging Amazon Echo – a hands free voice controlled speaker which is said to have sold 20 million units this year – can be said to be the fourth category.

There are also many areas where AI technology is being deployed where a special dedicated device is not needed, such as self-driving cars and the Advanced Driving Assistance System (ADAS). The finance industry is also using AI, areas like quantitative trading now use machine learning, for example. Medicine, law, human resources and the media are among the industries that are beginning to use AI. The Da Vinci surgical robot is able to perform simple, minimally invasive surgical tasks with high accuracy. In the future, medical technology is also expected to be used in long-term care; robots may be used to help rehabilitate patients and assist them in walking, for example. In the legal world, AI can be used to review simple contracts and record trials. When Microsoft was recruiting recently, it received 10,000 applications for new job postings and used an intelligent automated system to review and sort the applicants. Though AI cannot accurately find the one best applicant from a pool of 10,000 it did effectively screen and eliminate around 7,000 applicants. The media industry has also tried using AI to produce news stories; sports and financial news are said to be the easiest areas to cover in this way.

Opportunities for Entrepreneurs in AI

There are two areas of AI where entrepreneurs may find opportunities. One is in what is called critical applications, which includes things like automated driving and surgical robots. These applications require a high degree of accuracy, making it a difficult field to enter; the product cycle is long and it requires a large, long-term financial investment. The other opportunities are in non-critical applications, such as facial recognition, which require less accuracy. An example would be a robot that delivers your food order. It isn’t catastrophic if the robot gets your order wrong, trial and error and some human guidance can help improve the process over time.

In order to be competitive in this area, an entrepreneur needs to have a good understanding of the industry and its pain points in order to be able to propose targeted solutions. Then, as with most industries, you need the right engineering and production capabilities, cost control and supply chain management, and marketing skill.

When Doing AI, Go All In

There is no future in simply providing technology solutions. What is the basic law of an industry chain? When there are a lot of links but a company has a monopoly on one link, then the monopolist will eat up all the profits within the industry. For example in the PC industry chain there are many links, including hardware manufacturers, memory, CPUs, and operating systems, yet it’s clear that Windows and Intel make the lion’s share of the profits.

As the AI ecosystem grows, technical service providers should try their best to cover the whole stack, forming a closed loop of technology, products, business and data; only then can a company be valuable over the long term. If you only have technology, but no products, you cannot make a lot of money, and it will be difficult for you to form a data loop, because the field of artificial intelligence is dependent on data; the more data you have, the more accurate your algorithm will be. Apple and Tesla are very successful examples of doing the whole stack. So if you want to start a business in artificial intelligence, you must do the whole stack; that’s what industry players call artificial intelligence plus.”