AI in a Time of COVID-19

Pestilence is stalking the land, the shelves of Costco are emptying out, and the Fourth Horseman of the Apocalypse is up and riding.  Or is it just a serious case of the flu?

Whatever is happening biologically with this strain of coronavirus, “going viral” more than ever shares its original and virtual meanings.  We are now seeing attempts to manage the effects of a pandemic for the first time in our age of revitalized AI.  A substantial part of the response to the threat and reality of COVID-19 engages or is powered by artificial intelligence.  Here are four examples.

 

Control and Monitoring

China’s highly publicized response to the emergence of COVID-19 was the blunt instrument of total quarantine in the region around Wuhan, and other geospatial controls on travel and movement elsewhere.  This involved physical roadblocks and the deployment of wide and deep direct interaction with sick and well people alike, including constant taking of temperatures. According to an article posted today on Fortune Magazine’s website, “facial recognition cameras, phone tracking technology and voluntary registrations have all been deployed to monitor the flow of people and the possible transmission of disease.”     To what extent information about individuals is being retained on the government’s vast citizen information database is not yet known, but certainly doing so would be characteristic of China’s authoritarian policy.

As the world watched and waited, China monitored and reported the spread of cases through accounting practices that morphed over time, raising serious credibility questions as to the actual numbers of infected people and mortality rates.  It has long been recognized, however, that the collection and AI-powered analysis of social media posts, searches and chatbot questions provide an alternative  source for gauging public health that may be less open to state or political influence.

According to Prof. John Brownstein of Harvard Medical School, social media-based search systems like Harvard’s HealthMap are being used to monitor the global outbreak of COVID-19 through an unprecedented degree of data sharing and international collaboration.

 

Vaccine Development

On February 11-12, the World Health Organization (WHO) convened a meeting with scientists from biopharma companies, universities and regulatory agencies to create a roadmap for the coordinating the development and selection of vaccine candidates for advanced testing and related R&D.     At least 37 companies and academic groups, including 25 in China, have announced COVID-19 vaccine development programs.   Per the BioCentury article, “that work includes validating animal models to assess vaccine immunogenicity, efficacy and safety, and developing standardized reagents to evaluate vaccines.”  Such validation will undoubtedly involve algorithmic analysis and computer-based modelling.

Indeed, earlier in February, scientists in Australia reported the use of a set of algorithms they named “Smart Algorithms for Medical Discovery (“SAM”) to create for the first time solely from algorithmic modeling a “turbocharged” version of a flu vaccine.   And the complications of moving forward in a massively parallel vaccine effort will also involve AI-powered project management tools.

 

Computerized Trading in Capital Markets

When 4 trillion dollars is erased from the world’s stock markets in the fastest “correction” ever, one has to ask how much of that speed was due to algorithms that dominate trading in capital markets, and what are those algorithms  “learning” going forward about the financial implications of pandemic?  An article in Politico last Friday cited concerns of market regulators “that new practices and products [specifically high frequency trading] could increase market volatility and potentially amplify the pain of a market downturn.”  AI-based sorting of data inputs and resulting computer trading “judgments” of course is the steam that powers such trading practices.

But AI-powered computer modelling also has enabled the creation of new financial markets for investing in the risk of disasters like pandemics in the form of pandemic catastrophe bonds.  “Cat Bonds” are a means by which bond issuers, typically reinsurers or government agencies filling in for insurers in risk sectors that the private market has abandoned like California earthquake or Florida hurricane, issue bonds based on AI-driven powerful computer models.  These bonds are made possible by specialized cat bond modelling companies like AIR WorldWide demonstrating the low likelihood (often 6%) of the modelled risk occurring.  Whether the risk has “occurred” is pre-determined by objective triggers such as specific measures of disease.  The bonds pay enticing dividends well beyond the modelled risk percentage to buyers like hedge funds who are always on the lookout for new forms of diversity and who bet that the computer model of risk underlying the bonds is correct.

Such cat bonds were issued by underwriters several years ago to raise funds for the World Bank’s Pandemic Emergency Financing Facility, intended to help poor countries deal with the consequences of a severe pandemic. The bonds not only provide financial relief if they are triggered for payout, but the secondary markets in which they trade provide a strong signal of the potential for a disease outbreak to become a global pandemic.  This is because a trigger of the bonds will cause the loss of a bondholder’s entire investment.

As of last Friday, the riskiest of the WHO pandemic bonds were trading at between 60 and 70 cents on the dollar, signaling a significant likelihood that COVID-19 will soon be declared a global pandemic by WHO,  the investors in those bonds will lose their investment, and poor countries will benefit from the bond proceeds.

 

Disaster Demand Pricing

When disaster looms, price gouging is never far behind.  Certainly that has proven true with the spread to the West of COVID-19.  WIRED Magazine broke the story on February 25 of Amazon’s ethical challenges in half-heartedly applying its policies against price gouging on respiratory masks and hand sanitizer at amounts in many multiples of retail prices only a week or two ago.  Amazon’s Fair Pricing Policy prohibits listing items for significantly more than “recent prices offered on or off Amazon.”  But as WIRED notes, “the way Amazon exerts control over independent merchants has become a sensitive issue for the retail giant over the past year”  with global financial regulators. In some ways Amazon can’t win in market interventions.  To see how Amazon is currently balancing its interests, check out prices currently on offer for Purell – today worth its weight in quality cologne from some retailers.

So as in all aspects of life AI impacts,  we continue to see that it is simply a tool for good or ill.  It is up to human intelligence and ethical judgment to decide in which direction it will be deployed.

X