Opening Remarks at the Artificial Intelligence for Good Global Summit
Dr. Margaret Chan, Director-General of the World Health Organization
7 June 2017
Fellow UN agencies, experts in artificial intelligence, representatives of industry, ladies and gentlemen,
I welcome this opportunity to learn from the vast amount of technical expertise assembled in this room.
Market analysts predict that intelligent machines, programmed to think and reason like the human mind, will revolutionize health care in the very near future. In fact, proponents of the transformative power of artificial intelligence usually give two examples: self-driving cars and the delivery of health care.
This year’s influential Internet Trends Report, released last week in the USA, covers the effects of new technology on health care for the first time, again predicting a huge transformative impact.
Artificial intelligence is a new frontier for the health sector. As so often happens, the speed of technological advances has outpaced our ability to reflect these advances in sound public policies and address a number of ethical dilemmas.
Many questions do not yet have answers and we are not yet sure we know all the questions that need to be asked.
Much of the enthusiasm for the use of smart machines to improve health care reflects the perspectives of wealthy countries and well-resourced private companies. We need a broader perspective.
I find it wise to look at potential benefits, risks, and ethical dilemmas in the context of several worldwide trends that shape priority health needs.
Over the past decade, I have visited many countries where the majority of health facilities lack such basics as electricity and running water. I would be hard-pressed to sell these countries on the advantages of artificial intelligence when even standard machines for analysing patient samples or sterilizing equipment cannot run for want of electrical supplies.
Any discussion of the potential of smart machines to revolutionize the delivery of health care must be alert to these huge gaps in basic capacities.
At the same time, I have also observed the ubiquitous presence of smartphone even in the most resource-constrained settings. Schools may not have toilets or latrines. Children may not have shoes. But smart phones are ready to hand.
The traditional dichotomy between health conditions in rich and poor countries no longer holds. Health everywhere is being shaped by the same dominant forces, namely population ageing, rapid unplanned urbanization, and the globalized marketing of unhealthy products.
Under the pressure of these forces, chronic noncommunicable diseases have overtaken infectious diseases as the leading killers worldwide.
Diseases like heart disease, cancer, diabetes, and chronic respiratory diseases are profoundly shaped by human behaviours and the environments in which people make their lifestyle choices.
These are among the most democratic of all diseases, affecting all income groups in all places. They are also the most costly.
Could artificial intelligence help improve lifestyle choices? Could smart machines help consumers understand the meaning of food labels or interpret restaurant menu options? Could a smartphone app help people with diabetes maintain good metabolic control between visits to a doctor?
Moreover, the demands of long-term if not life-long treatment for chronic conditions have placed unsustainable pressure on an overloaded health workforce. The high-level Commission on health employment and economic growth estimates that the management of NCDs and conditions like dementia will require 40 million new health workers by 2030 in wealthy countries. In contrast, the developing world is expected to experience a shortfall of 18 million health workers.
The waves of populism and anti-globalization sentiment that are sweeping through some parts of the world are driven, in part, by technological advances that have eliminated many jobs, especially for the middle class.
Given the significant shortage of health workers, the application of artificial intelligence to health care could potentially reduce some of the burden on overloaded health staff. This is one advantage: revolutionary new technologies will certainly meet some resistance from the medical profession, but not, for the time being, because they threaten jobs.
Given the power of super computers and superchips to mine and organize huge amounts of data, it is easy to envision a number of applications in the health sector.
As we all know, health information is often messy and poorly structured. In many cases, it is systematically collected but not systematically analysed and used. Artificial intelligence can give that data a structure, and by detecting patterns, guide some medical decisions.
Supercomputers can accelerate the screening of novel molecules in the search for new drugs. They can speed up the reading and interpretation of results from radiographs, electrocardiograms, ultrasound and CT scans, and even the analysis of blood samples.
By reducing the likelihood of human errors, they can contribute to more precise diagnoses and predictions of patient prognoses, and to enhanced patient safety.
Other applications currently under development include personal use of smartphones to communicate symptoms and obtain a diagnosis from the cloud. Enthusiastic developers see this as a way to cut down health care costs by keeping the worried well from flooding clinics and emergency rooms.
For patients recovering from a stroke or an accident, developers have already introduced a system, involving sensor technology and the latest advances in cloud computing, that provides tailor-made physiotherapy that can be performed in homes. Immediate feedback scores the number of right and wrong movements. The cost of the system is estimated at one tenth of that for facility-based physiotherapy.
Ladies and gentlemen,
In the midst of all this exciting potential, I see several reasons for caution.
First, medical decisions are complex. They depend on context and values such as care and compassion. I doubt that a machine will ever be able to imitate genuine human compassion.
Second, machines can aid the work of doctors, organize, rationalize, and streamline the processes leading to a diagnosis or other medical decision, but artificial intelligence cannot replace doctors and nurses in their interactions with patients.
Third, we must consider the context and what it means for the lives of people. What good does it do to get an early diagnosis of skin or breast cancer if a country offers no opportunity for treatment, has no specialists or specialized facilities and equipment, or if the price of medicines is unaffordable for both patients and the health system?
What happens if a diagnosis by smartphone app misses a symptom that signals a severe underlying disease? Can you sue a machine for medical malpractice?
Medicines and medical devices are heavily regulated, and with good reason. Medical schools are accredited. Doctors and nurses are licensed to practice and are often required to undergo continuing education. How do you regulate a machine programmed to think like a human?
Regulatory issues must be solved before a new AI technology reaches the market. The reliability of wearable devices for monitoring cardiovascular performance is already being questioned. Medical history is replete with examples of technologies that were eventually rejected because they created a false sense of safety and security.
The mining of huge amounts of data raises serious issues of patient privacy and the sacrosanct confidentiality of medical records. This is another set of issues that must be addressed in advance.
Finally, we need to keep in mind that many developing countries do not have a great deal of health data to mine. These are countries that still do not have functioning information systems for civil registration and vital cause-of-death statistics.
In short, the potential of AI in health care is huge, but so is the need to take some precautions.