Artificial intelligence vs COVID-19: limitations, constraints and pitfalls
- PDF / 678,321 Bytes
- 5 Pages / 595.276 x 790.866 pts Page_size
- 55 Downloads / 216 Views
CURMUDGEON CORNER
Artificial intelligence vs COVID‑19: limitations, constraints and pitfalls Wim Naudé1 Received: 1 April 2020 / Accepted: 3 April 2020 © The Author(s) 2020
Abstract This paper provides an early evaluation of Artificial Intelligence (AI) against COVID-19. The main areas where AI can contribute to the fight against COVID-19 are discussed. It is concluded that AI has not yet been impactful against COVID19. Its use is hampered by a lack of data, and by too much data. Overcoming these constraints will require a careful balance between data privacy and public health, and rigorous human-AI interaction. It is unlikely that these will be addressed in time to be of much help during the present pandemic. In the meantime, extensive gathering of diagnostic data on who is infectious will be essential to save lives, train AI, and limit economic damages. Keywords COVID-19 · Data science · AI · Surveillance · Public health
1 Introduction The COVID-19 pandemic poses a number of challenges to the Artificial Intelligence (AI) Community. Among these challenges are “Can AI help track and predict the spread of the infection?”, “Can AI help in making diagnoses and prognoses?”, “Can it be used in the search for treatments and a vaccine?” and “Can it be used for social control?” This paper is an attempt to provide an early review of how AI have so far been contributing in this regard, and to note limitations, constraints, and pitfalls. These include a lack of data, too much (noisy and outlier) data, and growing tension between data privacy concerns and public health imperatives. To start out, let me discuss the actual and potential uses of AI in the fight against COVID-19.
2 Tracking and prediction AI can in principle be used to track and to predict how the COVID-19 disease will spread over time and space. In fact, an AI-based model of HealthMap, at Boston Children’s * Wim Naudé [email protected]‑aachen.de 1
Technology, Innovation, Entrepreneurship and Marketing, RWTH Aachen University, Kackertstrasse 7, 52072 Aachen, Germany
Hospital (USA), sounded one of the first alarms on 30 December 2019, around 30 minutes earlier than a scientist at the Program for Monitoring Emerging Diseases (PMED) issued an alert (see the discussion in Naudé 2020). For the further tracking and prediction of how COVID-19 will spread, however, AI has so far not been very useful. This is for a number of reasons. The first is that AI requires data on COVID-19 to train. An example of how this can be done is the case of the 2015 Zika- virus, whose spread was ex post predicted using a dynamic neural network (Akhtar et al. 2019). Because COVID-19 is different from Zika, or other infections, and because there are at the time of writing still not sufficient data to build AI models that can track and forecast its spread. Most of the growing number of publications reporting on using AI for diagnostic and predictive purposes so far tend to use small, possibly biased, and mostly Chinese-based samples, and have not been peer-reviewed. A number of promisi
Data Loading...
