Original article can be found here (source): Artificial Intelligence on Medium
AI and the Coronavirus: Part One
By Tej Kohli @MrTejKohli
If necessity really is the mother of invention, the coronavirus emergency is certainly exposing multiple areas of dire need. Over time, medical and other scientific innovations will surely emerge to limit the loss of life and protect humans against this new threat.
Those two words will provide little comfort for those currently struggling with the Covid-19 illness or fearful of its power. The world needs action right now to help mitigate, control and ultimately defeat this new threat. However, the new necessity of limiting the virus’s spread can also provide creative uses for knowledge and technologies that we already possess.
Take artificial intelligence. The concept officially became an academic discipline as long ago as 1955 and the 65 years since then have seen it divided into sub-fields and transformed by the Internet, robotics and the Internet of Things. Now, Ghost Data, a group of researchers based in Italy and the US, has utilised it to provide some support for Italian authorities fighting the crisis.
Automatic for the People
Specifically, the group, which has previously applied AI to countering terrorism, online counterfeiting and digital propaganda, wanted to help the Italian government understand how citizens were evading the public lockdown implemented to prevent the virus spreading.
Harvesting images and videos from 552,000 social media profiles and 504,000 public Instagram stories, they anonymised the data, removing personal information, blurring faces and decreasing geo-location precision. They then used other relevant data to determine regional locations before processing the information through visual AI technology created by LogoGrab, an Irish software firm that specialises in identifying logos in images and videos.
Object and scene detection technology was used to identify images of people outdoors or in public indoor spaces, such as bars and restaurants, as well as inside vehicles. The final data set was then meshed with Italy’s strict lockdown laws to identify cases of possible avoidance.
The results revealed that 40% of contraventions took place in cities, 26% were on beaches and 17% were in parks. Some 16% of violators were walking or trekking, 7% were playing sports and 6% were driving vehicles. Some 40% were members of larger gatherings.
The data also reveals the worst-behaving regions (Campania and Lombardy), while future analysis could include time-stamped trends, enabling even more granular detail to help the authorities police transgressions. The exercise can obviously be repeated to assess any variance in compliance and its correlation with infection rates.
Although the exercise did include manual reviews and no attempt was made to prosecute based on the data, it provides a further insight into how the innovations of AI ventures can free humankind from mundane tasks.
Further work may be required to refine the “people-centric” methodology and there will be myriad other potential applications too. Could the model be adapted to monitor anti-social behaviour such as grocery stockpiling? Would it help identify the preponderance of blackmarket trade in hand sanitisers? Could it even locate the hidden stacks of toilet rolls?
Of course, it has such potential. While libertarians may fret, this kind of intervention may not even be as intrusive as CCTV imaging. Questions do remain about the potential of AI to play Big Brother and how far society should go down this road. Where lives are at stake and a killer virus is on the loose, however, such matters should fade into insignificance.
With police and local authority inspectors under pressure due to the virus, society needs all the help it can get from artificial intelligence right now. How it is deployed to prevent this crisis worsening may have lasting, beneficial consequences for us all.
Tej Kohli is an investor who backs growth-stage artificial intelligence and robotics ventures through the Kohli Ventures investment vehicle.