Pandemic helped democratise AI, ML and data science, says Techvantage CEO

When it started, there was very little knowledge about Covid-19. It pushed mankind to harness every ounce of available technological innovation and ingenuity to fight it. AI and ML have played a key role in understanding and addressing the crisis. AI truly has become one of humans’ best friends and has helped in detection, prevention, response of and recovery from the disease. The big challenge was that most health systems could not handle the unanticipated patient load. There was a felt need to prioritise patients for treatment.

We helped a few clients to quickly re-engineer their human-intensive, vulnerable processes and replace them with automated processes powered by AI and ML – for instance, an auto-dispatch customer service ticketing system, prospecting algorithms to prioritise end customers with high chance of conversion, contactless customer service with chatbots, ML-enabled chatbots for contactless screening of Covid symptoms, and video-enabled, auto-dispatch of insurance claims.

The timing is just right to work on these. There are several studies, some of them anecdotal, that suggest pre-existing medical conditions can potentially influence the prognosis and post-Covid quality of life. There are still challenges with availability and quality of data. There are definitely more Covid-related public datasets available than before, but they are distributed across heterogenous sources and, for the large part, unstructured. Data about pre-existing medical conditions is harder to access. In order to build something that will be useful for ‘patient level prediction’, AI and medical professionals as well as others should collaborate to create partnerships and build an ecosystem where reliable data can be sourced from and the created models tested and improved.

Setting up of an institute of this size and scale in full compliance with regulatory guidelines is by no means an easy task. It is likely to take time before we will have enough data to create mitigation strategies. A confluence of various stand-alone databases (example: healthcare data with immigration data) can potentially stop a lot of transmissions early on. Researchers have leveraged AI to discover a treatment that could stop the outbreak. There are deep learning models that can predict old and new drugs or treatments. Having data and capability to build an AI platform to make use of it to aid drug design and drug repurposing is only the first step. Partnership with capable drug research and manufacturing firms that offers subject matter expertise is essential.

No data is usually available ready for consumption by an AI platform. The reason is that many of these data capture systems were not created with the expectation that the data will someday be consumed by AI platforms. It is usually the responsibility of the data analyst or data scientist to get it ready for consumption. In India, the data pertaining to Covid-19 comes from heterogenous sources. Vaccine compliance data is on the CO-WIN platform whereas the test data is with ICMR. It is quite possible that these source systems may not also hold data with the best quality. They may also be in different formats. It is imperative that we analyse available data in order to prepare the country for a potential future outbreak.

We need to promote responsible use of personal data. It is likely that there will be an increasing trend towards the use of invasive collection, processing and sharing of personal health and behavioural data for targeted monitoring of individuals. While these measures are essential, governments should ensure these tools are implemented responsibly. There should also be a resolve to cease or reverse exceptional uses of data after the crisis. Policymakers must ensure alignment to sound principles (example: OECD AI principles) that respect human rights and privacy. The systems should be transparent, secure, safe and the developers should remain accountable.