You might have heard about those ATMs that use facial recognition instead of cards and PIN numbers for authentication. You might also have seen on the news a smart security algorithm that helps police identify suspects and cracks criminal cases. Artificial intelligence (AI), the wiz behind these advanced technologies, is permeating our daily lives—everything from financial services to public safety to healthcare and transportation.
YITU Technology, one of China’s front-running AI startups, has developed solutions that help solve real-world problems. YITU now has the ability to enable accurate facial recognition with a large database of over 1 billion faces in just one second, and their technology has in fact assisted Chinese law enforcement in criminal investigations.
In January, the company launched its first international office in Singapore and expects to commit more resources to AI research in Southeast Asia and other parts of the world.
To get a peek behind the curtain of AI research and development, TechNode spoke to YITU Technology’s AI research scientist Dr. Wu Shuang and visited YITU’s headquarters in Shanghai.
Located in a high-rise in Shanghai’s Hongqiao business district, the head office has the aesthetics of a Silicon Valley tech startup with a special fondness for glass windows and open workspaces.
Instead of swiping ID cards or filling out visitor sign-in sheets, employees simply scan their faces by standing before a facial recognition system—a perfect demonstration of YITU’s technology. Walking through the working area, it was almost impossible not to notice the strong presence of security cameras watching from every corner.
At one of the well-lit open spaces, a large screen displayed a real-time map showing the exact time and location of each individual in the office—demonstrating how a public safety solution based on facial recognition and location tracking technologies can be implemented in public spaces.
YITU’s AI-based products have been successfully incorporated in a number of smart city solutions, and already in use in banking, healthcare, transportation and public safety settings.
In 2015, YITU teamed up with Aliyun to build the Big Data Real-time Cloud System for Guizhou Traffic Police. In the same year, YITU implemented its facial recognition technology to realize cardless ATM withdraw, which has been rolled out across China Merchant Bank’s network of 2,000 ATMs.
Read More: A year in constant review: China’s surveillance breakthroughs in 2017
Revolutionizing medical diagnosis and clinical research
AI is much needed in healthcare, where a myriad of opportunities has surfaced as the technology advances. YITU has implemented AI technology in medical diagnosis and clinical research, Dr. Wu Shuang told TechNode.
Based on medical records, the startup has developed a number of patient-facing products including CT-based lung cancer detection and screening, x-ray based child bone age prediction, and pediatric diagnosis system — all have a direct impact on patient lives.
Their information retrieval system is revolutionizing the way doctors conduct clinical research—an area where the main demand is retrieving and filtering large amounts of data efficiently. AI technology saves doctors and medical researchers the trouble of manually sifting through medical records. But, building a comprehensive medical record search engine is more challenging than it sounds, Wu said. Training the system to understand medical expertise and jargon is a strenuous task. The company currently works with 30 of the top hundred hospitals in China.
Too good to be true?
While AI certainly helps in medicine—assisting professionals trained in its use and application—can AI and humans really communicate in a real-world setting?
“Chatbots are a very hot topic right now. Most chatbots they don’t keep track of what they have been talking about, that shows you how far we are from solving the problem.”
The challenging part is to train the system to understand the human thought process. “In most scenarios, you’re asking the algorithm to give you an answer and that is it, you don’t ask a lot of follow-up questions.” Google search engine, for example, treats search terms as independent queries—when a user types in a query the system produces a list of relevant results, but when the user keys in the second query, the system immediately forgets about the first query. “That is because it is very difficult [for machines] to follow this kind of train of thought people have.”
Wu said currently in real-world settings, AI is still not up to the job. “With a deeper understanding of the situation and possibilities of how the situation can change, there are models that can do this sort of thing. But I will say it is still far from being very practical.”
Read more: Almost 80% of Chinese concerned about AI threat to privacy, 32% already feel a threat to their work
Democratizing AI
When it comes to attracting new AI talent, Wu said encouragingly, “it’s not a plus or minus for anybody, because everyone is facing the same problem.” However, “a lot of information is public: the papers are public and basically all the research progress made are public.” Right now, nobody and no big companies can say they dominate or control most IP or anything in that respect, Wu said encouragingly. Big companies can’t not publish their research as a way to dominate the field of AI.
AI powerhouses like Google, Microsoft, Facebook, Baidu and many others have unleashed their immense technological power — openly publishing research and released software on open source platforms for anyone to download and use.
AI research and development is global. Wu said much has been happening in non-first world countries as governments begin to realize the impact of AI. “Just like any new technology, if it turns out to be practical and impactful… individuals, academia, corporations, and countries are all going to try to make the right approach to benefit from it.”
—With contribution from Jeriel Tan.