Artificial Intelligence during a pandemic: The <scp>COVID</scp>‐19 example

Abstract

Artificial intelligence (AI) is transforming our lifestyle intending to mimic human intelligence by a computer/machine in solving various issues. Initially, AI was designed to overcome simpler problems like winning a chess game, language recognition, image retrieval, among others. With the technological advancements, AI is getting increasingly sophisticated at doing what humans do, but more efficiently, rapidly, and at a lower cost in solving complex problems. AI in healthcare provides an upper hand undoubtedly over traditional analytics and clinical decision-making techniques. Machine learning (ML) algorithms, a subset of AI, can detect patterns from huge complex datasets to become more precise and accurate as they interact with training data, allowing humans to gain unprecedented insights into early detection of diseases, drug discovery, diagnostics, healthcare processes, treatment variability, and patient outcomes.1 But how effective are the AI algorithms during a disease outbreak or for that matter a pandemic? After 2000, the pandemics are testing the AI's ability to handle extreme events. The two major factors affecting AI algorithms include the availability of historical and real-time data and high computational power. The different roles played by AI during pandemics are early warning and alerts, prediction and detection of outbreak of diseases, real-time disease monitoring worldwide, analysis and visualisation of spreading trends, prediction of infection rate and infection trend, rapid decision-making to identify the effective treatments, study and analysis of the pathogens, and drug discovery. All these are executed at a greater speed with AI. WHO and CDC (United States) are receiving data of several diseases and situations occurring across the world. With modern computer architecture and internet, all these data can be accessed in real-time by different institutes to develop an autonomous or collaborative AI model to handle various tasks. In addition to the official data, AI can gather information from news outlets, forums, healthcare reports, travel data, social media posts, and others in multiple languages across the world by using natural language processing (NLP) techniques and flag their priority. Several terabytes of data which includes patients' case history, geographical events, and social media posts about a new pneumonia are processed at a rapid rate with high-performance computing to predict the possible outbreak of a pandemic.1-3 Most importantly unsupervised ML can identify its own pattern from the noise (historical and real-time data) rather than the training it on a preselected dataset, thus giving a wider possibility and new behaviour. An AI model trained to predict a particular disease can be retrained on the new data of a new or different disease. There are also few AI models that are a hit and miss due to lack of historical training data. Though AI has not completely evolved to overcome a pandemic, but the role of AI is noticeably high during COVID-19 when compared to that of previous pandemics and is rightly used as a tool complementing the human intelligence. The authors declare no conflicts of interest.

Ähnliche Arbeiten