Chad Mills

Artificial Intelligence: fighting coronavirus behind the scenes

Artificial Intelligence is playing a critical role in the fight against coronavirus, though it can be hard to understand how exactly it’s contributing in our world full of sensational headlines. I’ll help to cut through this noise.

But first, what are others saying?

Some stories claim AI is going to completely solve the problem, with headlines like Why AI might be the most effective weapon we have to fight COVID-19 or even AI identified coronavirus before it spread.

Others have a completely opposite perspective, saying Coronavirus reveals Limits of AI Health Tools or AI could help with the next pandemic—but not this one.

This entire spectrum, which includes most articles on this topic, presumes that AI is a magical tool that is supposed to solve all problems. The responses vary in whether or not that result is being achieved. It’s not the content in these stories that’s wrong, but rather this basic perspective itself.

The role of AI in society

Before getting into how AI is playing a part in the fight against COVID-19, it’s helpful to reflect on how the technology is impacting our daily lives. It will turn out, unsurprisingly, that the impact it’s having on efforts to fight the virus is similar.

For the most part, we go about our daily lives making the key decisions that shape our lives: what jobs to take, who to marry, where to live, what friends we have, how much time to spend with family, etc. We choose our own values.

Our social fabric includes tools that help us achieve these values, and many of these tools include a component based on artificial intelligence.

Facebook helps us stay in touch with friends and loved ones, and AI helps us see the most interesting posts so we don’t need to scroll through pages of promotional posts from the local pizzeria we like.

When we need to find information, we type a few words into Google and it points the way to the answers we’re looking for.

Netflix enables us to choose what shows to watch, but also offers suggestions that help us find good shows we might not have otherwise found.

In each of these cases, we’re in charge but the AI algorithms that help us are in the background, making these tools more useful and enabling us to enjoy life more than we could have without them. The role AI is playing in fighting the new coronavirus is similar.

Fighting COVID-19: understanding the virus

How do we even know this sickness is a coronavirus like SARS or MERS? AI played an important supporting role here.

Scientists were able to sequence the genome of this novel coronavirus in record time. This enables them to understand what it is, track its spread, and discover how to treat it.

The first time the human genome was sequenced, it took 13 years and nearly a billion dollars, completed in 2003. Today, it takes 1-2 days and several thousand dollars to completely sequence another human’s genome.

Tools are available that make genomic sequencing relatively straightforward. The genetic material is broken down into small fragments. Fluorescent-colored genetic building blocks are added and the fragments are copied with this new material; the sequence of colors in these copies reveals the structure of each fragment.

Millions of overlapping fragments are read, and computer programs can stitch these back together into an overall picture of the genome.

As a recent research paper indicates, “[i]n biological research machine learning algorithms are part of nearly every analytical process.” This process of sequencing COVID-19 was no different.

The first sequence produced, for example, used the sequencing technology described above. Researchers used an approach widely used in statistics and artificial intelligence (maximum likelihood) to show this new virus is related to a known coronavirus from bats, not too different than SARS.

AI, here, is not overarching magic that solves the problem on its own. Most of the actual work is at the biological and chemical levels. But tools based on artificial intelligence can quickly analyze the result of this work and dramatically speed up the process, making things possible that wouldn’t be feasible without AI. This is very similar to the critical supporting role artificial intelligence plays in our daily lives.

What does this mean for stopping the threat?

The COVID-19 pandemic is growing at an alarming rate. Even with understanding the genome of the virus in record time, treatments and vaccines won’t be available fast enough to solve the problem without major disruptions to everyday life.

In the United States, a vaccine is believed to be 12-24 months away since the development process must include extensive testing over a significant period of time to ensure the treatment is safe and effective. Whether these limits will remain is an open question, but these are not issues AI is in a position to solve.

Artificial intelligence is enabling the response to proceed faster than it ever has historically, but it’s still not as fast as the virus. Also, advances in science and medicine, such as the chemical processes involved in sequencing DNA, are enormous contributors to the sped-up process. AI helps, but it’s not all AI.

There’s no doubt that artificial intelligence is saving numerous lives by making the response faster and more effective. Saying it’s solving the problem, however, would be an overstatement.

How else is AI helping?

With this context of the supporting role AI plays in mind, we can turn to the more sensational ways the technology is helping the fight against COVID-19. Keep in mind that in each of these cases, what AI is accomplishing is part of the effort.

Early detection of pandemic risk

A quick response is critical to reducing the impact of and being prepared for a new infectious disease. Companies using artificial intelligence are making it easier to spot these trends before they become widely-known.

By translating text from foreign languages, monitoring forums where doctors communicate, tracking trends on social media, and more, systems can look for troubling indicators of new infectious diseases that could become a major health crisis.

HealthMap, a product of Boston Children’s Hospital, spotted local physicians sharing their concerns on an online forum in Chinese and raised an alarm before anyone else outside of China. This beat the Chinese government and the WHO at detecting the threat.

That said, this system identified the risk as three out of five. This could be enough if the system is just flagging them for human review, but it clearly isn’t accurate enough to stand on its own.

BlueDot, a system that also predicted the spread of Ebola and the Zika virus, also flagged this as a SARS-like problem before humans produced warnings.

AI is able to bring in data from a wide variety of sources and across languages, playing a supporting role in detecting risks—even if the output still requires human assessment and response.

Understanding the spread

Other approaches can help as the threat spreads. Metabiota tracks travel patterns and assesses the risk of where a threat like this will spread. They were able to quickly incorporate travel restrictions into their model, successfully predicting Italy, Iran, and the United States were major risks for early spread.

Smart devices, like Kinsa’s AI-driven thermometers, store health data in the cloud similar to Fitbit and can enable tracking population-level trends in aggregate to show how a threat is actually spreading.

As the virus has mutated and scientists have sequenced the genome around the globe, Graphen has analyzed the evolution of over 370 strains, showing which parts are most susceptible to mutation and how it has spread.


South Korea has perhaps handled this outbreak best, with the key being widespread testing.

South Korean company Seegene made a diagnostic test in less than two weeks using AI—and they did so from WHO data, without even having access to the virus. They didn’t even wait until the virus came to South Korea before they started building the test.

Using AI enabled them to build tests at low-cost for a range of rare conditions. This one “went viral” and they were ahead of the game; it wasn’t a superhuman effort on their part, but rather the unique workflow and cost profile of artificial intelligence solutions that made this feasible.

Seegene is producing enough to test a million people each day, and beyond South Korea they’re also shipping test kits to Italy, Germany, and many other countries.

In-hospital testing

Beyond the standard tests, chest CT scans are a critical way to quickly diagnose cases when someone who poatentially has COVID-19 walks into a hospital.

There have been a wide range of AI-based solutions to diagnose the unique patterns of COVID-19 from this imaging. Infervision, Alibaba, Ping, and hospital and university researchers have all developed high-accuracy tests that evaluate the imagery about 60 times faster than humans.

Meanwhile, there’s also an AI-based test that runs in 40-60 minutes without needing imagery. Or an app can help people assess their risk and manage how they respond. Another tool tries to predict people who may have coronavirus before even running a test.

In a world where the supplies needed to execute a test are running out and imaging machines themselves take time to run, this does not mean AI has solved the testing problem. It does help, including with bottlenecks like analyzing CT scans, but there are other bottlenecks it can’t solve.


Artificial intelligence can also help doctors understand how well patients are likely to fare as their symptoms progress.

One test based on bloodwork is able to predict survival outcomes with 90% accuracy. Another can identify patients likely to take a turn for the worse.

These types of predictions are imperfect, but they can help determine the best course of treatment and even encourage earlier intervention to support those likely to need additional care.

Another company, Biofourmis, has a monitoring program for people in quarantine. By wearing a device that tracks oxygen levels, temperature, etc., doctors can better understand the way the sickness develops in the earlier stages or even identify early people who may need additional treatment.

Reducing the spread

There are several widespread ways that AI has been used to help reduce the spread of the new coronavirus.

First, in some places temperature sensors are used in population centers to detect people with fevers who should be separated from the crowd. China in particular has aggressively pursued this strategy.

Multiple companies are combining this with facial recognition to identify people who may be infected. These temperature sensors have even been placed on police helmets, at border crossings, and on drones.

Sensors are also used to listen for coughs and count them, not identifying particular people at risk but giving a general risk assessment in a population or area.

Another way to reduce transmission is to remove human interaction that could lead to disease transmission.

Drones deliver medical supplies between disease control and a hospital in China. Robots deliver food and sterilize packages. Other robots communicate with medical staff, take vital measurements, clean hospital rooms, and disinfects outside areas.

A very different approach to reducing spread is by using wearable devices. One such device detects when its wearer is reaching to touch his or her face and produces a warning; this reduces one of the risk factors for catching the virus.

Understanding COVID-19

I previously described in detail how scientists were able to sequence the genome of the new coronavirus and this relied on AI.

Nevertheless, this is only the beginning. Once scientists understand the structure of the virus, further understanding can enable treatments.

In particular, researchers at DeepMind have used the genetic sequence to predict the structure of proteins associated with the virus. This is different than observing them empirically, but can speed up such experiments by helping scientists know what to look for.

Meanwhile, Microsoft Research, the National Library of Medicine, and the Allen Institute for AI have collated tens of thousands of research papers related to coronaviruses and this virus in particular, making them available to artificial intelligence researchers.

Since it would be difficult for a human to read all of these papers, using artificial intelligence may facilitate finding patterns that might otherwise be difficult to spot.

Developing treatments

Artificial intelligence is also hard at work helping medical researchers to find ways to treat or prevent COVID-19 quickly.

BenevolentAI and Deargen are able to use AI to predict whether existing drugs could help to fight coronavirus.

This is critical since one of the things that makes drug development take so long is the process of ensuring the medicine is safe for wide-scale consumption. Medications that are known to be safe could be made available immediately if they help.

Other companies are using AI to help find new treatments.

Insilico Medicine simulated tens of thousands of molecules over four days to see which might stop the virus from replicating, and narrowed it down to a small list worty trying. They’ve already synthesized two of the top seven to test in the next couple weeks.

Iktos is leveraging AI to design new molecules to fight the virus, and has partnered with SRI International to generate real-world molecules from these designs for experiments. They hope to take drug discovery down from two years to six months.

Moderna Therapeutics uses a combination of AI and researchers to find a vaccine, and is already planning pre-clinical trials in April in Seattle. Healx also encourages such collaborations, using human researchers to help guide the algorithms in finding a vaccine.

To support all this ongoing research, some large tech companies have made AI platforms available for researchers to use for free.

Finding cures and treatments is amazing, but testing them out in the physical world requires substantial effort. AI again is a significant contributor but not the entire solution.

Information quality

With many people in quarantine or ordered to shelter-in-place in a drastic effort to reduce virus transmission, fear is rampant and getting accurate information is critical.

Chatbots from Tencent (WeChat), Microsoft and the CDC, and CloudMedx share reliable information personalized to the individual, even helping people figure out whether they should be tested.

Meanwhile, the large tech platforms employ large armies of contractors to moderate content on their platforms, which includes fighting against medical misinformation.

With much of this work involving sensitive data, many of these contractors are not allowed to work from home. When offices close due to fears of spreading coronavirus, this leaves the platforms vulnerable. Facebook, Google, and Twitter are all increasing their reliance on AI to solve these problems in these challenging circumstances.

Behind the scenes

AI is not an overarching intelligence solving problems like this new virus all on its own. Instead, the technology plays an important supporting role in many of the activities pursued to fight the virus.

This list could be endless, so these are just a few examples.

Ant Financial is using artificial intelligence to speed up medical claim processing. In addition to reducing face-to-face interactions between patients and staff, it also saves healthcare workers time in a crisis where their time is already stretched thin.

There’s a major shortage of masks and other personal protective equipment for healthcare workers. Sonovia uses artificial intelligence to infuse fabric with nanoparticles to produce anti-viral properties; they then use this fabric to produce masks.

In some areas, AI is used to help ambulances make their way through traffic so they can get patients in urgent need to the hospital right away.

There are many small roles like these that AI plays in supporting the response to COVID-19.


Artificial intelligence plays an important supporting role in our daily lives, and it’s no different in how it’s helping fight the new coronavirus.

AI is helping with a wide range of problems: providing early warning signals of a pandemic, predicting the spread, making tests for a new virus quickly available, reducing the spread, understanding the virus, developing treatments, helping people get reliable information, and much more.

In each of these cases, though, artificial intelligence is not a magical system that solves the problem completely. AI can’t, on its own, swab a patient’s nose, test a new treatment, or even produce mass quantities of a new treatment when it becomes available.

What it can do, however, is make many steps in these processes faster and consume less resources. While not a silver bullet, AI is providing help we desperately need.

About author View all posts Author website


Chad currently leads applied research, ML engineering, and computational linguistics teams at Grammarly.

He's previously led ML and data science teams at companies large and small, including working on News Feed at Facebook and on Windows and Outlook at Microsoft.

Leave a Reply

Your email address will not be published. Required fields are marked *