4 ways that AI has helped during the pandemic

Not many people realise, but it was actually an AI-automated system that first alerted the wider world to COVID-19. On December 30 2019, the HealthMap algorithm at Boston Children’s Hospital first flagged the potential danger, and another AI model, the Canada-based ‘Blue Dot’, issued a similar alert the next day. Both bots had been using natural language processing (NLP) to scan social media and online news articles to inform public health teams. Very quickly, AI researchers were able to pinpoint the top 20 destination cities for passengers arriving from Wuhan. In hindsight, it wasn’t enough, but it shows the potential of AI to monitor for global outbreaks. Now AI researchers are teaming with tech companies to build automated tracking systems to better mine the vast amount of online data.

When you can predict something, you can prepare for it. That’s the goal of several AI researchers who are building models to protect vulnerable populations from COVID-19. One start-up, Closedloop, is using public health data to create an open-sourced ‘COVID Vulnerability Index’ – a predictive model that can identify those most at-risk from severe COVID complications. Public health teams will be able to access this model to offer targeted support where it’s needed most: contactless supply drop-offs, public health information, extra medical resources etc. In fact, scaling up efficiently is one area where AI really shines. In March 2020, Canadian AI developers worked with the Johns Hopkins Bloomberg Medical School of Public Health to create an algorithm that predicted how many hospital beds would be needed in various US states, and the optimal duration of stay-at-home orders.

One of the toughest manual tasks for public health teams during the pandemic has been contact tracing. Partly this is down to the inherent unreliability and forgetfulness of interviewees, but mostly it’s because of the exponential growth factor: if one COVID patient infected three others, and each of them infects three others, the potential pool of close contacts quickly explodes beyond the ability of health teams to control. This is where AI can help. In 2020, Victoria purchased an AI platform from IBM to assist their contact tracing team. The model helped identify close contacts faster, and even potential super spreaders and hot-spot suburbs. Still, there’s a long way to go. Fully automated tracking has been rolled out in dozens of apps around the world (with varying degrees of success) and it looks like the best solution is still a mix of manual, on-the-ground contact tracing and AI support.

AI has been a huge tool for COVID researchers and health teams around the world, often in unexpected ways. AWS, for example, has launched CORD-19 Search, built on the Allen Institute for AI’s open dataset of more than 128,000 research papers, to quickly allow scientists to look-up questions about COVID-19. Specific, technical questions like “When is the salivary viral load highest?” Other developers have repurposed pneumonia medical imaging algorithms to rapidly diagnose COVID-related lung scans (UC San Diego Health is on the cutting-edge of this tech). AI is even helping in the race for COVID vaccines and anti-viral drugs. In 2020, a UK company, BenevolentAI, used an AI drug discovery platform to figure out which compounds stood the best chance against COVID-19 (they quickly found that Baricitinib, a rheumatoid arthritis drug, was the best candidate for COVID patients, and it rapidly began entering US clinical trials).

Of course, while AI is a powerful too, it’s still a tool. And that means it can be used wisely or poorly. Artificial intelligence offers humanity a weapon against COVID-19, but it’s up to us to use it effectively.

Interested in how AI can drive business value and shape the future of work? Check out our range of AI short courses below.