AI in the Age of COVID-19

Original article can be found here (source): Artificial Intelligence on Medium

AI in the Age of COVID-19

Oren Etzioni and Nicole DeCario

Photo by Markus Spiske on Unsplash

COVID-19 has been declared a global pandemic, and AI is playing an increasingly vital role in everything from virus detection to treatment and prevention. In this article, we identify very recent developments and future potential, but also concerns about the societal impact of AI and robots being developed to fight infectious disease.

AI is serving three notable roles in the current crisis: reducing the spread of infection, treating those who are already ill, and preventing future outbreaks. Since the virus is transmitted by close contact with an infected person, hospitals are reducing human interaction by deploying robots to disinfect rooms and provide telehealth consultations. In Wuhan, an entire hospital ward is now staffed with robots. Over 3,400 Chinese healthcare workers have contracted the virus. The robots work 24/7 to help protect medical staff from infection.

Researchers are actively working to find effective treatments for COVID-19, which has already claimed over 21,000 lives worldwide. The AI community has come together to support researchers’ efforts by providing datasets and creating algorithms that will uncover valuable information and save critical time. Google DeepMind rapidly trained its existing AI system to map the structure of the novel coronavirus and has released their findings publicly. Understanding the protein structure of a virus is a complicated, lengthy process; DeepMind’s work could reduce that time by months.

Recognizing the problem of information overload for researchers, the Allen Institute for AI launched the free Semantic Scholar literature search engine in 2015 to “cut through the clutter”. Semantic Scholar leverages AI techniques including natural language processing and computer vision methods to improve the academic search experience in four key ways. First, we provide AI-based methods to rank relevant papers to help researchers prioritize their reading. Second, we connect papers with relevant datasets, software, news articles, presentation slides, videos, and even tweets to streamline the process of mining insights and results from the scientific literature. Third, we provide insights into the connections between papers by analyzing each paper’s references. Finally, we have a personalized research “feed” (analogous to a Twitter or Facebook feed) that identifies which recent papers (including pre-prints) a researcher would want to see based on their interests.

In response to the COVID-19 outbreak, Semantic Scholar has applied its capabilities to the relevant academic literature and created the COVID-19 Open Research Dataset (CORD-19), a corpus of tens of thousands of papers about COVID-19 and related coronaviruses. This resource represents the most extensive machine-readable coronavirus literature collection available for data and text mining to date. It was released on March 16, 2020 on the Semantic Scholar website and created in collaboration with the Chan Zuckerberg Initiative, Cold Spring Harbor Laboratory, Georgetown University’s Center for Security and Emerging Technology, Microsoft Research, and the National Library of Medicine of the National Institutes of Health. Its contents will continue to be updated as new research becomes available.

AI has been part of the COVID-19 story since day 1. AI startup BlueDot created software that uses natural language processing and machine learning to review data from hundreds of thousands of sources to locate, track, and predict infectious disease spread. The software detected a cluster of unusual pneumonia cases in Wuhan in late December and accurately predicted where the virus might spread. AI is also being used to map the spread of infection in real time, diagnose infections, predict mortality risk, and assist with triage.

The potential for future AI innovation in disease response and prevention is staggering.

The potential for future AI innovation in disease response and prevention is staggering. Microbots are being developed to swim through bodily fluids to deliver drugs or other medical relief in a highly targeted way. Robotic arms can be used to perform mouth swabs, use a stethoscope, or perform ultrasounds. Drug development is already occurring more rapidly. A new medicine to treat obsessive-compulsive disorder was “invented” by AI in just 12-months. Typically, it takes five years for such a drug to reach human trials.

With these rapid advances comes a deep responsibility to consider the implications of these new technologies. While deploying robots to take over an entire hospital ward has obvious benefits in a crisis, we must consider what happens to the staff they’ve replaced. Job loss due to automation is already a hotly debated issue. The use of surveillance technology is another high-profile topic that has found a place in the coronavirus fight. With fever as a key symptom of COVID-19, fever detection systems have been deployed in public spaces in an effort to halt the spread of the virus. The privacy and ethical consequences of such uses should not be overlooked.

AI and robotics have already contributed to the fight against COVID-19 and have the potential to be scaled and improved in the future. However, policymakers ought to consider whether we are opening a Pandora’s box that will negatively impact privacy and displace millions of workers. After all, COVID-19 will fade with time, but the technologies we deploy may be here to stay.