Skip to Content

New post-lockdown use cases for Artificial Intelligence in China

Capgemini
April 17, 2020

In our previous blog post of this series, we shared how during the lockdown Chinese retailers and brands brought to life a new way of shopping, that can be thought of as Touchless Retail. In this blog, we look at some of the new AI-based tools being used in China post the lockdown.

In the recent ‘Report on Data and Smart Applications for Epidemic Prevention and Control’from the China Academy of Information and Communications Technology’  these use cases were categorized into three main areas, as indicated here:

Movement of people

Geo-localization

Whether it is to track your incoming meal on a delivery platform or to locate the bike nearest to you on a bike-sharing app, the number of mobile apps using geo-localization has exploded in urban China over the past decade. In the context of COVID-19, one of the most widely used applications during the recent lockdown was provided by Baidu Maps.The function displayed the infected areas in real time, and alerted users of a potential risk before they intended to travel.
With permission granted to access their GPS data, users were given information on infected locations: both nearby and for specific locations searched. The app detailed the exact distance to infected areas and sent alerts about crowded locations. On top of the GPS data provided by telecom networks, the health data was mainly provided by official sources, such as local Medical and Health Departments. Developed in just a few weeks, the additional function in Baidu Maps was immediately adopted and, by the end of February, it counted nearly 100 million searches.

The interface of Baidu Map shows infected areas nearby

In addition to guiding individuals, geo-localization data was used to model the general public’s movement and to simulate post-lockdown scenarios in order to plan the path to recovery.

Fourth Paradigm, a Chinese AI startup, developed the Intelligent Disease Control Platform to model deconfinement options and anticipate the risk of disease spread across regions, as well as the potential peak of the outbreak. The simulation platform used data, such as people’s movement history, number of suspected infections, confirmed and recovered cases as inputs. It could then help to simulate post-lockdown scenarios for cities and provide options for reopening traffic routes.

After being a part of governmental outbreak control, Fourth Paradigm raised a total of $230 million in its latest Series C and C+ funding rounds on April 2nd and now approaches $2 billion in valuation.

Bluetooth

Elsewhere in Asia, Singapore opted for big data technologies to help during the crisis, including a government-introduced application called TraceTogether. This used Bluetooth records to identify people who had been in direct contact with any newly infected patient identified.

Users could download this app in their smartphones, which then exchanged short-distance Bluetooth signals when users were close to each other. Information being exchanged included: a timestamp, Bluetooth signal strength, the phone’s model, and a temporary identifier. This contact data was then stored for 21 days for later investigation when needed. If someone was diagnosed with COVID-19, Singapore’s Health Ministry could access the data and identify those who had been in close contact with the patient.

Health condition reporting

Health Code

In China, AI was also used to report public health status and grant access to public locations. Using the Health-Code technology, the information was carried in an individual health QR-code available on a smartphone, which could be scanned to allow access to secured places.

Implemented in over 200 cities (most of the largest ones), Health Codes were widely used by Chinese citizens to access public areas. They worked as a digital pass: a color code indicated the user’s potential risk of infection, based on their declared health status and history of movements. Launched by local governments and tech companies like Alibaba and Tencent, the Health Code was generated by individual citizens, who had to indicate their personal data on initiation (including their ID number and health status). The code was initiated to “Green” for people without an infection record and history of travel in a high-risk area. Users were required to regularly scan QR codes at building entrances in order to refresh their color status.When travelling to a place recorded as risky, a user’s health-code turned to “Red” and they were quarantined.
By early March, there were more than 1.6 billion Tencent Health Codes generated covering more than 900 million citizens. In cities like Shanghai, the system allowed the public to move freely within a couple of weeks of lockdown ending.

The QR code for scanning to query personal health status at the entrance of office building

Face recognition and temperature control

Facial Recognition Thermometers were widely used in China’s public areas. These thermometers had been “trained” to detect when someone was not wearing a mask and could alert security. Simultaneously the system measured the temperature with an infrared thermometer and alerted if it was above the threshold. Behind the system were deep learning algorithms trained to focus on key features around eyes and eyebrows. One of the most recognized systems was developed by Baidu, which was used over 32,000 times, with 190 suspected cases found in one month after its launch in late January at Beijing Qinghe Railway Station.

The facial recognition screen with alert of people without mask

AI-assisted medical support.

Robotic health status check and advice

Dialog robots were developed to track people’s health condition and respond with health advice to people’s enquiries 24/7. Using contact and travel records, an AI tool identified people to be monitored. The robot then arranged direct calls to collect health information and follow up on highly suspect cases. Meanwhile, conversations were recorded into a database, from which further analysis of health trends could be made. The robots could also answer general public enquiries, advise on planning clinical visits and educate on disease prevention.

AI-assisted diagnosis

Lastly, AI-based technologies assisted in providing medical support and advice. During the outbreak, almost every province experienced a massive shortage of medical capacity, especially in the highly infected areas, which was the major bottleneck in scaling down confirmed cases.AI was used in combination with computer-assisted tomography (CAT) to accelerate and increase the accuracy of the diagnosis phase.
Alibaba launched a ‘CAT+AI’ Diagnosis Assistant to provide a COVID-19 diagnosis in less than 20 seconds after the CAT images were uploaded in the cloud. Based on image recognition and deep learning technologies, this AI tool achieved an accuracy of 96% on diagnosis, including examination on cases confirmation, level of lung infection and stage of disease. As of March 31st, this AI medical tool had been adopted in over 170 hospitals for 340,000 potential patients. Other countries like Japan have also adopted the technology to help the diagnosis phase.

Innovation and AI

The accelerated widespread adoption of AI technologies in China and elsewhere in Asia is indicative of a broader global trend, for example with self-reporting health apps becoming the norm. As the above shows, new technologies are coming onto the market that have the potential to shape healthcare provision and monitoring in the years ahead.

At the same time, ethical challenges arise with the use of AI as it pertains to personal information. While the current crisis has made people more amenable to sharing their details for health and wellbeing purposes, organizations and governments have a moral obligation to leverage personal data and technology ethically going forward. Getting the right balance between usage and the need to monitor the likelihood of the virus reappearing in the future will be an ongoing challenge.

The content in this blog is a personal analysis by the authors based on observations made locally and may not always be compliant with regulations in other geographies


Co-authors

Carol Wang

Consultant

Capgemini Invent

Angie Zhang

Consultant

Capgemini Invent

Lorraine Zhu

Consultant

Capgemini Invent