“In 35 years of practicing medicine, I’ve never seen anything like it.” The words came from a friend who is an ER doctor who was talking with me over the phone. He’s been treating COVID-19 patients and is now in quarantine himself after being exposed.
He’s treated pneumonia, colds, flu, and other conditions, but COVID-19 just doesn’t act like any of these more common illnesses.
SEE: Coronavirus: Critical IT policies and tools every business needs (TechRepublic Premium)
It seems that artificial intelligence (AI) and machine learning (ML) have important roles to play as physicians, hospitals, pharmaceutical companies, and others directly engaged in the coronavirus fight try to find rapid therapeutic and containment answers.
How can AI and ML help fight this disease?
AI and ML can track disease spread trends that identify where healthcare needs are most acute and where the virus is likely to spread, helping facilitate early intervention. They can do this by analyzing past spread patterns, learning from these, and combining that knowledge with insights into how spread is perpetuated and where it is likely to go next, based on population flows and other factors.
AI can also match COVID-19 symptoms and treatments with other therapeutic treatments that have been administered in similar symptom scenarios in order to identify best treatments for the disease while we await a vaccine. At the same time, AI and ML can analyze millions of pharmaceutical elements and formulas, and speed time to a vaccine.
Managing the risks
On the flip side, those using AI and ML for COVID-19 must also manage the risks.
One risk is the 95% de facto standard of “correctness” for algorithms that defines when an AI application is ready for production. Do we strictly adhere to that standard, especially when data sample size might be limited? Do we even know what 95% “correctness” for COVID-19 looks like, since we have little experience with this virus? If we don’t adhere, can we accept something like 85% “correctness” to speed solutions to market? If you are a non-healthcare company impacted by COVID-19, such as insurance, do you use AI to predict losses? Can AI be used to predict supply chain shortages and plan for them?
SEE: 4 tips for developing better data algorithms (TechRepublic)
These are a few of the questions organizations around the world want answered—but in a
worldwide pandemic, there is another question that nations and companies must answer:
Can they collaborate with others on AI to combat coronavirus?
On March 24, 2020, Costa Rica asked the World Health Organization (WHO) to create a voluntary intellectual property pool to develop affordable and universally available COVID-19 drugs. Initially there was a U.S. pushback, and it is possible that other countries and pharma firms feel the same.
The better news for global and inter-company cooperation on COVID is that the WHO is launching a multi-country clinical trial for potential coronavirus therapies. China shared research which indicated that men are more adversely affected by COVID-19 than women or children. The open source community is working on COVID data algorithms in its “Just One Giant Lab” initiative, aimed at improving methods for testing and detecting the virus.
For any company implementing AI, the COVID-19 crisis has a clear message: AI is no longer an experimental or expendable technology. The coronavirus crisis has demonstrated that AI is a mission-critical technology, and corporate IT plans should be redrafted to reflect this, no matter how their AI is being deployed.