Coronavirus: How we can harness AI to save lives

In 1918, the H1N1 virus, which would eventually become known as the Spanish flu, ripped through the world, killing some 50 million people, despite the fact that the global population at the time was relatively static and isolated compared to the highly connected world of today.

Science at the time was poorly equipped to tackle it. Only decades later would scientists identify that the “cytokine storm”, an expression to describe the overreaction of the body’s immune system that Covid-19 has made familiar today, was also the killer in the 1918 pandemic. Fast forward a century and the world faces a new pandemic, but now our science and technology are better able to respond with research and development to limit the toll of the Sars-CoV-2 virus (popularly known as the novel coronavirus) and the Covid-19 disease it causes.

Digital technologies are clearly central contributors to this new capability. Some make it possible for us to work from home and socialise remotely, and others enable coherent responses to the crisis. But it is Artificial Intelligence or AI that is powering the global search for solutions. With its ability to analyse and learn from the vast and complex data sets that this pandemic rapidly generates, it is key to this search.

AI is, of course, central to the vision of Sheikh Mohamed bin Zayed, Crown Prince of Abu Dhabi and Deputy Supreme Commander of the Armed Forces. His vision inspired the launch of the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and its key research theme: the use of AI for the benefit of society.

There is currently no greater societal need than dealing with this ongoing pandemic.

We don’t need to look far for practical examples.

In Abu Dhabi, Seha, the government health organisation that owns and operates public hospitals – with its 60 outpatient clinics, 12 hospitals throughout the emirate and 14 drive-through testing centres across the country – is using AI to provide the most effective response to Covid-19.

Seha’s AI modelling has also shown that the coronavirus infection rates in Abu Dhabi are unlikely to accelerate over the next few weeks due to the effective measures in place.

Around the world, work is progressing at breakneck speed – to manage the spread of the disease; to identify treatments; to develop an effective vaccine and to ensure people are reliably informed.

Trace, Track and Test

As governments around the world seek to manage the spread of the virus, re-establish economic activity and allow some return of social life, many will rely on TTT: test, trace and track.

Singapore, China and South Korea adopted mobile tracking and tracing solutions early, and the use of such solutions will increasingly be deployed to safely loosen current lockdowns.

Google and Apple have committed to work jointly with governments and health agencies to help develop a shared platform that provides basic tracking and tracing technology.

But privacy concerns loom large. Trace.ai, a start-up that utilises AI to identify people using 2,000 characteristics – but crucially not facial recognition – and claims 98 per cent accuracy, may answer some concerns.

The two main tests are for the presence of the disease itself and a test for antibodies to show that an individual has been exposed and may have developed some immunity.

Concerns about the selectivity, availability and turnaround times of current tests provide an incentive to search for other approaches, too. Hospitals using algorithms to examine CT scans are reportedly reaching more than 90 per cent accuracy in diagnosing Covid-19.

Microsoft and Seattle-based Adaptive Biotechnologies are aiming to build a practical technology to map and decode the human immune system by coupling AI and machine learning with recent breakthroughs in biotechnology.

Their stated ambitious goal is to create a universal blood test that reads a person’s immune system to look for molecular indicators of a wide variety of diseases and disorders, including Covid-19.

Treatments

In the UK, BenevolentAI aims to use artificial intelligence to identify potentially effective medicines, with one such drug – baricitinib – currently entering clinical trials with Covid-19 patients. Sometimes a cocktail of existing drugs can target a new disease.

Award-winning AI company Healx uses its platform to test drug combinations: 4,000 drugs means eight million pairs and more than 10 billion triples. In Singapore and China, academics lead the project Identif.AI, harnessing AI to find optimal drug combinations and doses to fight the virus.

Outcomes of Covid-19 vary from negligible to disastrous, with some worst affected having pre-existing health conditions.

Organ damage has been found in some patients following severe Covid-19 disease. A company that I chair, Perspectum, has launched a study to better understand the impact of this disease, using AI analysis of MRI organ scans.

Vaccines

The race to discover a vaccine is unprecedented, with more than 70 vaccine development programmes under way by late April and a few entering clinical trials, including a programme led by my colleagues at the University of Oxford that has recently begun potentially breakthrough trials with humans.

Using AI and cloud computing, Flinders University in Australia has developed a potential Covid-19 vaccine that prevents the “spike protein”, which is used by the virus to invade human cells, from binding to the ACE2 receptor on human cells.

Japan-based NEC Corporation has used AI prediction platforms to design blueprints for potent and widely effective Sars-CoV-2 vaccines. It has already produced an analysis of results.

Two of AI’s key roles in this search are to suggest components of a vaccine by understanding viral protein structures and to help medical researchers scour tens of thousands of relevant research papers at unprecedented speed.

In short order, teams at the Allen Institute for AI, Google DeepMind, and elsewhere have created AI tools and shared data sets and research results freely with the global scientific community.

Public information

On March 9 in Boston, Partners HealthCare opened a hotline to answer questions and concerns about Covid-19. It was quickly overwhelmed.

The solution was an AI-based chatbox with a straightforward user interface, presenting callers with questions based on content from the US Centres for Disease Control and Prevention.

A similar system deployed in Seattle dealt with an astonishing 40,000 patients in its first week.

Conversely, problems also centre on proliferating disinformation as well as misinformation, with potentially harmful consequences for public health and effective crisis communication.

The World Health Organisation has said that false claims "are spreading faster than the virus", terming it an "infodemic of planetary proportions". Large technology companies use AI extensively to fight fake news and the tools are improving all the time.

The challenges the world is facing today are unprecedented and how we as a global community respond to this crisis will be studied for decades and possibly centuries to come.

Thankfully, we don’t have to imagine tackling the Covid-19 crisis without modern tools and technologies.

And while AI doesn’t promise us all the answers, it does add an important toolkit to help the world’s research and science community, governments and health providers to save lives and reduce the impact of this 21st century pandemic.

Professor Sir Michael Brady is the interim president of Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)