This post was originally published by the Council of Europe
Artificial intelligence (AI) is being used as a tool to support the fight against the viral pandemic that has affected the entire world since the beginning of 2020.
The press and the scientific community are echoing the high hopes that data science and AI can be used to confront the coronavirus (D. Yakobovitch, How to fight the Coronavirus with AI and Data Science, Medium, 15 February 2020) and “fill in the blanks” still left by science (G. Ratnam, Can AI Fill in the Blanks About Coronavirus? Think So Experts, Government Technology, 17 March 2020).
China, the first epicentre of this disease and renowned for its technological advance in this field, has tried to use this to its real advantage. Its uses seem to have included support for measures restricting the movement of populations, forecasting the evolution of disease outbreaks and research for the development of a vaccine or treatment. With regard to the latter aspect, AI has been used to speed up genome sequencing, make faster diagnoses, carry out scanner analyses or, more occasionally, handle maintenance and delivery robots (A. Chun, In a time of coronavirus, China’s investment in AI is paying off in a big way, South China Morning post, 18 March 2020).
Its contributions, which are also undeniable in terms of organising better access to scientific publications or supporting research, does not eliminate the need for clinical test phases nor does it replace human expertise entirely. The structural issues encountered by health infrastructures in this crisis situation are not due to technological solutions but to the organisation of health services, which should be able to prevent such situations occurring (Article 11 of the European Social Charter). Emergency measures using technological solutions, including AI, should also be assessed at the end of the crisis. Those that infringe on individual freedoms should not be trivialised on the pretext of a better protection of the population. The provisions of Convention 108+ should in particular continue to be applied.
The contribution of artificial intelligence to the search for a cure
The first application of AI expected in the face of a health crisis is certainly the assistance to researchers to find a vaccine able to protect caregivers and contain the pandemic. Biomedicine and research rely on a large number of techniques, among which the various applications of computer science and statistics have already been making a contribution for a long time. The use of AI is therefore part of this continuity.
The predictions of the virus structure generated by AI have already saved scientists months of experimentation. AI seems to have provided significant support in this sense, even if it is limited due to so-called “continuous” rules and infinite combinatorics for the study of protein folding. The American start-up Moderna has distinguished itself by its mastery of a biotechnology based on messenger ribonucleic acid (mRNA) for which the study of protein folding is essential. It has managed to significantly reduce the time required to develop a prototype vaccine testable on humans thanks to the support of bioinformatics, of which AI is an integral part.
Similarly, Chinese technology giant Baidu, in partnership with Oregon State University and the University of Rochester, published its Linearfold prediction algorithm in February 2020 to study the same protein folding. This algorithm is much faster than traditional algorithms in predicting the structure of a virus’ secondary ribonucleic acid (RNA) and provides scientists with additional information on how viruses spread. The prediction of the secondary structure of the RNA sequence of Covid-19 would thus have been calculated by Linearfold in 27 seconds instead of 55 minutes (Baidu, How Baidu is bringing AI to the fight against coronavirus, MIT Technology Review, 11 March 2020). DeepMind, a subsidiary of Google’s parent company, Alphabet, has also shared its predictions of coronavirus protein structures with its AlphaFold AI system (J. Jumper, K. Tunyasuvunakool, P. Kohli, D. Hassabis et al, Computational predictions of protein structures associated with COVID-19, DeepMind, 5 March 2020). IBM, Amazon, Google and Microsoft have also provided the computing power of their servers to the US authorities to process very large datasets in epidemiology, bioinformatics and molecular modelling (F. Lardinois, IBM, Amazon, Google and Microsoft partner with White House to provide compute resources for COVID-19 research, Techcrunch, 22 March 2020).
Artificial intelligence, a driving force for knowledge sharing
In the United States, the White House Office of Science and Technology Policy met with technology companies and major research groups on 11 March 2020, to determine how AI tools could be used to, among other things, screen the thousands of research papers published worldwide on the pandemic (A. Boyle, White House seeks the aid of tech titans to combat coronavirus and misinformation, GeekWire, March 11, 2020).
Indeed, in the weeks following the appearance of the new coronavirus in Wuhan, China, in December 2019, nearly 2,000 research papers were published on the effects of this new virus, on possible treatments, and on the dynamics of the pandemic. This influx of scientific literature naturally reflects the eagerness of researchers to deal with this major health crisis, but it also represents a real challenge for anyone hoping to exploit it.
Microsoft Research, the National Library of Medicine and the Allen Institute for AI (AI2) therefore presented their work on 16 March 2020, which consisted of collecting and preparing more than 29,000 documents relating to the new virus and the broader family of coronaviruses, 13,000 of which were processed so that computers could read the underlying data, as well as information on authors and their affiliations. Kaggle, a Google subsidiary and platform that usually organises data science competitions, created challenges around 10 key questions related to the coronavirus. These questions range from risk factors and non-drug treatments to the genetic properties of the virus and vaccine development efforts. The project also involves the Chan Zuckerberg Initiative (named after Facebook founder Mark Zuckerberg and his wife Priscilla Chan) and Georgetown University’s Center for Security and Emerging Technologies (W. Knight, Researchers Will Deploy AI to Better Understand Coronavirus, Wired, March 17, 2020).
Artificial intelligence, observer and predictor of the evolution of the pandemic
The Canadian company BlueDot is credited with the early detection of the virus using an AI and its ability to continuously review over 100 data sets, such as news, airline ticket sales, demographics, climate data and animal populations. BlueDot detected what was then considered an outbreak of pneumonia in Wuhan, China on 31 December 2019 and identified the cities most likely to experience this outbreak (C. Stieg, How this Canadian start-up spotted coronavirus before everyone else knew about it, CNBC, March 3, 2020).
A team of researchers working with the Boston Children’s Hospital has also developed an AI to track the spread of the coronavirus. Called HealthMap, the system integrates data from Google searches, social media and blogs, as well as discussion forums: sources of information that epidemiologists do not usually use, but which are useful for identifying the first signs of an outbreak and assessing public response (A. Johnson, How Artificial Intelligence is Aiding the fight Against Coronavirus, Datainnovation, March 13, 2020).
The International Research Centre for Artificial Intelligence (IRCAI) in Slovenia, under the auspices of UNESCO, has launched an “intelligent” media watch on coronavirus called Corona Virus Media Watch which provides updates on global and national news based on a selection of media with open online information. The tool, also developed with the support of the OECD and the Event Registry information extraction technology, is presented as a useful source of information for policy makers, the media and the public to observe emerging trends related to Covid-19 in their countries and around the world.
Artificial intelligence to assist healthcare personnel
For their part, two Chinese companies have developed AI-based coronavirus diagnostic software. The Beijing-based start-up Infervision has trained its software to detect lung problems using computed tomography (CT) scans. Originally used to diagnose lung cancer, the software can also detect pneumonia associated with respiratory diseases such as coronavirus. At least 34 Chinese hospitals are reported to have used this technology to help them screen 32,000 suspected cases (T. Simonite, Chinese Hospitals Deploy AI to Help Diagnose Covid-19, Wired, February 26, 2020).
The Alibaba DAMO Academy, the research arm of the Chinese company Alibaba, has also trained an AI system to recognise coronaviruses with an accuracy claimed to be 96%. According to the company, the system could process the 300 to 400 scans needed to diagnose a coronavirus in 20 to 30 seconds, whereas the same operation would usually take an experienced doctor 10 to 15 minutes. The system is said to have helped at least 26 Chinese hospitals to review more than 30,000 cases (C. Li, How DAMO Academy’s AI System Detects Coronavirus Cases, Alizila, March 10, 2020).
In South Korea, AI is reported to have helped reduce the time needed to design testing kits based on the genetic make-up of the virus to a few weeks, when it would normally take two to three months. The biotech company Seegene used its automated test development system to develop the test kit and distribute it widely. Large-scale testing is indeed crucial to overcome containment measures and this testing policy seems to have contributed to the relative control of the pandemic in this country, which has equipped 118 medical establishments with this device and tested more than 230,000 people (I. Watson, S. Jeong, J. Hollingsworth, T. Booth, How this South Korean company created coronavirus test kits in three weeks, CNN World, March 13, 2020).
Artificial intelligence as a tool for population control
The example set by Singapore in its control of epidemic risks, with the support of technology, is certainly unique and difficult to export because of the social acceptance of restrictive safety measures: issue of a containment order for populations at risk, verification of compliance with the measures by mobile phone and geolocation, random home checks (K. Vaswani, Coronavirus: The detectives racing to contain the virus in Singapore, BBC News, 19 March 2020). AI has been quite widely used in support of such mass surveillance policies as in China, where devices have been used to measure temperature and recognize individuals or to equip law enforcement agencies with “smart” helmets capable of flagging individuals with high body temperature. Facial recognition devices have, however, experienced difficulties due to the wearing of surgical masks, leading one company to attempt to circumvent this difficulty since many services in China now rely on this technology, including state services for surveillance measures. Hanvon thus claims to have created a device to increase the recognition rate of wearers of surgical masks to 95% (M. Pollard, Even mask-wearers can be ID’d, China facial recognition firm says, Reuters, 9 March 2020). In Israel, a plan to use individual telephone follow-up to warn users not to mix with people potentially carrying the virus has been developed (A. Laurent, COVID-19: States use geolocalisation to know who respects containment, Usebk & Rica, 20 March 2020 – in French only). In South Korea, an alert transferred to the health authorities is triggered when people do not comply with the isolation period, for example by being in a crowded place such as on public transport or a shopping centre (Ibid.). In Taiwan, a mobile phone is given to infected persons and records their GPS location so that police can track their movements and ensure that they do not move away from their place of confinement (Ibid.). In Italy, a company has also developed a smartphone application that can be used to trace the itinerary of a person infected with the virus and warn people who have had contact with him or her. According to the designer, privacy would be guaranteed, as the application would not reveal phone numbers or personal data (E. Tebano, Coronavirus, pronta la app italiana per tracciare i contagi: ‘Così possiamo fermare l’epidemia’, Corriere della Sera, 18 March 2020) In Lombardy, telephone operators have made available data concerning the movement of mobile phones from one telephone terminal to another (M. Pennisi, Coronavirus, come funzionano il controllo delle celle e il tracciamento dei contagi. Il Garante: «Non bisogna improvvisare», Corriere della Sera, 20 March 2020).
In the United States, tension can be perceived between guaranteeing individual rights and protecting collective interests during this health crisis. Thus, the GAFAM have at their disposal in the United States information which would be extremely valuable in times of crisis: an immense amount of data on the American population. Larry Brilliant, an epidemiologist and executive director of Google.org, claims that he can “change the face of public health” and believes that “few things in life are more important than the question of whether major technologies are too powerful, but a pandemic is undoubtedly one of them” (N. Scola, Big Tech faces a ‘Big Brother’ trap on coronavirus, POLITICO, 18 March 2020). The U.S. government has therefore asked these companies to have access to aggregated and anonymous data, especially on mobile phones, in order to fight the spread of the virus (T. Romm, E. Dwoskin, C. Timberg, U.S. government, tech industry discussing ways to use smartphone location data to combat coronavirus, The Washington Post, March 18, 2020). However, these companies have been cautious in view of the legal risk and potential image damage (S. Overly, White House seeks Silicon Valley help battling coronavirus, POLITICO, 11 March 2020). Data regulation would likely have helped frame the public-private dialogue and determine what types of emergencies should be subject to the collective interest over individual rights (as well as the conditions and guarantees of such a mechanism), but Congress has made no progress in the last two years on such a law.
Finally, attempts at misinformation have proliferated on social networks and the Internet. Whether it concerns the virus itself, the way it spreads or the means to fight its effects, many rumours have circulated (“Fake news” and disinformation about the SARS-CoV2 coronavirus, INSERM, 19 February 2020). AI is a technology already used with some effectiveness by platforms to fight against inappropriate content. UNICEF adopted a statement on 9 March 2020 on misinformation about the coronavirus in which it intends to “actively take steps to provide accurate information about the virus by working with the World Health Organization, government authorities and online partners such as Facebook, Instagram, LinkedIn and TikTok, to ensure that accurate information and advice is available, as well as by taking steps to inform the public when inaccurate information appears”. The enactment of restrictive measures in Council of Europe member States to avoid fuelling public concern is also envisaged. However, the Council of Europe Committee of Experts on the Media Environment and Media Reform (MSI-REF) underlined in a statement of 21 March 2020 that “the crisis situation should not be used as a pretext to restrict public access to information. Nor should States introduce restrictions on media freedom beyond the limits allowed by Article 10 of the European Convention on Human Rights”. The Committee also highlights that “member States, together with all media actors, should strive to ensure an environment conducive to quality journalism”.
Artificial intelligence: an evaluation of its use in the aftermath of a crisis
Digital technology, including information technology and AI, are therefore proving to be important tools to help build a coordinated response to this pandemic. The multiple uses also illustrate the limits of what can currently be achieved by this very technology, which we cannot expect to compensate for structural difficulties such as those experienced by many health care institutions around the world. The search for efficiency and cost reduction in hospitals, often supported by information technology, should not reduce the quality of services or compromise universal access to care, even in exceptional circumstances.
It should be recalled that Article 11 of the European Social Charter (ratified by 34 of the 47 member States of the Council of Europe) establishes a right to health protection which commits the signatories “to take, either directly or in co-operation with public and private organisations, appropriate measures designed in particular to : 1°) to eliminate, as far as possible, the causes of ill-health; 2°) to provide consultation and education services for the improvement of health and the development of a sense of individual responsibility for health; 3°) to prevent, as far as possible, epidemic, endemic and other diseases, as well as accidents.”
Finally, it should be possible to evaluate the emergency measures taken at the end of the crisis in order to identify the benefits and issues encountered by the use of digital tools and AI. In particular, the temporary measures of control and mass monitoring of the population by this technology should not be trivialized nor become permanent (Y. N. Harari, Yuval Noah Harari: the world after coronavirus, The Financial Times, 20 March 2020).
Standards relating to data protection, such as Convention 108(+) of the Council of Europe, must still be applied fully and under all circumstances: whether it be the use of biometric data, geolocalisation, facial recognition or the use of health data. Use of emergency measures should be carried out in full consultation with data protection authorities and respect the dignity and the private life of the users. The different biases of the various types of surveillance operations should be considered, as these may cause significant discrimination (A.F. Cahn, John Veiszlemlein, COVID-19 tracking data and surveillance risks are more dangerous than their rewards, NBC News, 19 March 2020).
Original article published by the Council of Europe