Artificial intelligence plays multiple roles in tackling Covid-19

The Covid-19 pandemic response can draw on technological capabilities that were unavailable even a few years ago. One of those, artificial intelligence (AI), is playing a particularly vital role.

In early August, Russia shocked the world by announcing that it had approved the first vaccine for Covid-19. Experts have questioned the safety of the vaccine, which has not yet entered the final phase of clinical trials, but ‘Sputnik V’ may yet cement its place in history. If it doesn’t, there will be no shortage of contenders to take its place. More than 40 vaccine candidates are in trials worldwide.

This is a remarkably fast response to a virus first identified in December last year. The wealth of potential vaccines shows what can happen when the world focuses effort, expertise and resources on one problem, but they are also evidence of the technological leaps of recent years.

The compute and data storage capacity that can be deployed to tackle Covid-19 is way beyond what was available even in the early 2000s during the SARS outbreak. One area where technology has really advanced is in AI, which is central to the global pandemic response. Here are a couple of key examples...

AI virus response
London-based Kuano is one company that is turning its AI know-how to the pandemic. The company’s platform was originally developed to design new cancer treatments but is now being applied to Covid-19. One challenge with a new disease is the lack of information on how it works. Kuano deals with that by using drug and virus interactions for a known disease to help identify possible treatments for the emerging one.

Of course, the speed with which we learned about how the virus works was itself only possible because of AI. One way to gain information about a virus is through ’sequencing’ - the process of analysing its genetic material, which in the case of a coronavirus is a molecule called RNA. AI accelerated this process by predicting the virus’s structure and, particularly, how it might mutate. Baidu, the Chinese technology giant, used AI to cut the time taken to predict parts of the Covid-19 RNA sequence from 55 minutes to 27 seconds.

But the applications for AI go way beyond searching for a vaccine. In the weeks after Covid-19 was first discovered, 2,000 research papers were published about the virus. Tens of thousands more have been published since. AI excels at finding patterns in quantities of data that would swamp a human brain, so it is being used to sift through the research to spot things such as risk factors and non-drug treatments.

Beyond medical applications
Researchers in China have developed an algorithm that can analyse chest CT scan images and distinguish ordinary pneumonia from Covid-19 in seconds - which can be a vital advantage in isolating patients and slowing the spread of the disease. However, some uses of AI are more controversial than others. In China, infrared cameras are being used to scan crowds at airports and railway stations for people with high temperatures. Facial recognition is also reportedly being used to spot people who have disobeyed quarantine orders.

This is just a taste of the ways that AI and machine learning are being used in the Covid-19 response. None of them would be possible without significant computing power, which is one reason why high-performance computing (HPC) is so vital. The Covid-19 HPC Consortium, which includes IBM, BP and Google, as well as the Massachusetts Institute of Technology, NASA and the Swiss National Supercomputing Centre, has been coordinating pandemic-related projects.

In May, UK Research and Innovation (UKRI) joined the Consortium, adding “more than 20 Petaflops of high-performance computing capability to the global effort to address the coronavirus crisis”. This will help to accelerate research even more.

The Russians may or may not have developed the world’s first coronavirus vaccine but in all likelihood multiple vaccines will be needed. In the coming months more candidates will complete their trial processes. Those that are successful will owe much to AI and HPC.