Technology approach(es) used to catalyse investment:Other
Artificial Intelligence (AI) is technology that powers machines using human-like intelligence. AI-enabled machines can mimic humans, automate manual tasks and learn like humans. These machines are increasingly being used alongside biometrics. Together these systems are being used during disease outbreaks and pandemics to detect people displaying specific attributes, such as high temperatures to indicate fever. These systems can be used at a multitude of places where people congregate – including railway stations, hospitals and municipal buildings.
These systems scan about 200 people per minute, calculating their forehead temperature from up to three metres away. They can accurately detect temperature even when people are wearing masks, hats, or covering their faces. The system will generate an alert when it detects a high temperature that will prompt staff to carry out secondary checks.
Biometric technologies of today are typically used to authenticate and identify individuals (e.g. the passenger screening process at an airport or the fingerprint-scanning button of a smartphone). Now, in response to major country-wide or global disease pandemics such as Coronavirus (COVID-19), governments are deploying such technologies to identify infected persons and curtail the spread of disease. Several devices have been trialled across China and Singapore, targeting built up areas including public transport, in response to COVID-19.
With the global outbreak of COVID-19, countries are looking for unique ways to utilise technology to minimise the spread of the disease. To avoid enforcing mass quarantine on their citizens, governments are seeking ways to enable their citizens to continue moving around their regions to maintain as much normality to life as possible.
The initial means of achieving this was to implement manual thermometer testing in common areas for mass congregation of people, such as railway stations. This led to major queues in accessing these areas due to the time-consuming nature of the process. Such devices can only measure two to three people per minute and put front-line testers at risk of contracting the disease.
Alongside the increasing movement to automation and the growing sophistication of AI human intelligence, repetitive and time-consuming tasks will increasingly be undertaken by AI. What's more, AI-powered systems exhibit human intelligence and learn with time, which indicates that these machines can eventually carry out critical-thinking jobs and take decisions by themselves. For example, Baidu Maps is providing information on infected cases through data to track population migration and predict the spread of the pandemic in China.
Improving efficiency and reducing costs:
Enhancing economic, social and environmental value:
Legislation and regulation: Local authorities and transport agencies should develop a response plan for secondary testing and the subsequent processes to be followed when an alarm is triggered. These processes should reduce the potential for spread to passers-by and staff and should minimise anxiety. Data privacy legislation will need to be considered, which will vary by country. It is key that governments communicate with their citizens to explain these technologies and their benefits.
Transition of workforce capabilities: To effectively develop and implement these solutions, authorities will need to add pandemic requirements to their business case framework as well as digital programming and engineering support to translate those pandemic requirements into solution specifications. This will require inputs from health specialists and researchers.
Risk: One of the biggest problems with scanning for temperature or fever conditions is how far away people are from the device. Taking ambient temperature into account is also necessary to avoid faulting the detection. If people come from outside, they will appear hotter.
Mitigation: Further modification of the software could be required to enable it to consider changes in temperature based on distance. The AI system can learn what the compensation should be as the person approaches nearer and nearer and how to measure ambient temperature and take that into consideration.
Risk: Due to the growing complexity of machine learning, and the potential for human biases to influence the system, it can be difficult to understand how AI systems produce their results. It can be difficult to determine who is held responsible for when the system outputs go wrong: Is it the responsibility of the developer, tester or product manager?
Mitigation: Organizations must include internal and external checks to ensure equitable application across all participants and ensure that data and algorithms minimize discriminatory bias and avoid pitfalls introduced by humans during the coding process, to ensure there are no unintended or unfair consequences for users. Organizations should also make algorithms, attributes and correlations open to inspection so that participants can understand how their data is being used and how decisions are made.
Safety and (Cyber)security risk
Risk: There is a risk that users will reject the technology due to perceived or real threat that that data will be used for other purposes, or that there is a cybersecurity risk. People are particularly reluctant to share their health-related information.
Mitigation: Data privacy and cybersecurity considerations vary from country to country. This technology is designed to log the volume of traffic, timestamps, and temperatures and send these to a cloud-based system that can report the rate of traffic and number of fevers detected. The use case is not designed to capture other specific personal data and the relevance of the movements’ analysis and the protection of people’s identity should not be compromised by its processing in the application. The application should also be protected using the same proven and robust techniques and technologies commonly employed by high traffic commercial websites and includes data encryption, specific workflow to process and validate the data integrity and restricted user access.
Governments can also educate their citizens on the use of the data and prove that it will not be used to track citizens beyond the specific use case.
Example: Baidu's AI Tool, China
Implementation: Baidu introduced their systems at transport hubs in response to a call to improve temperature monitoring in the city in response to COVID-19. Baidu’s system can detect the temperature of moving masked people with a margin of error of 0.05°C.
Timeframe: The tool was developed from an existing Baidu tool and therefore was able to be developed quickly in response to the pandemic.
Example: DJI Temperature Screening Drone
Implementation: DJI adapted their existing drones, designed to monitor temperature fluctuation in industrial environments, to accurately measure fever in humans remotely. The drone is fitted with a FLIR Lepton thermal micro camera.A cotton swab was used to improve the cameras accuracy for human use. The drones were also used to spray disinfectant on to infrastructure (e.g. roads).
Cost: DJI pledged USD 1.5 million in aid to help contain the COVID-19 outbreak.
Timeframe: Minimal time required to develop, as DJI repurposed their existing drones to meet the public health requirements of the pandemic e.g. the existing Agras series of agricultural spraying drones was adapted to spray disinfectant in potentially affected areas.
Example: iThermo Tool, Singapore
Implementation: The Integrated Health Information Systems (IHiS), partnered with local healthcare AI startup KroniKare to pilot iThermo:an AI-powered temperature screening tool used to identify people showing symptoms of fever, in response to COVID-19.
Timeframe: KroniKare ramped up production to ensure 100 units of the device would be developed within a month of the pilot being agreed (February 2020).