INTERVIEW. “The obstacles to AI are more human than technical” – Sciences et Avenir

After posting Health and artificial intelligence with Cédric Villani in 2018, Bernard Nordlinger and him edited a new book, devoted more precisely to the applications of AI in the medical field. In Medicine and Artificial Intelligence, published on March 24, 2022 by CNRS, more than sixty experts presented the main challenges that the use of algorithms will pose in medical practice. During a symposium at the Academy of Medicine on May 3, 2022, Bernard Nordlinger, head of the digestive general surgery and oncology department at Ambroise-Paré, researcher and member of the Academy, answered our questions. .

Bernard Nordlinger. © College of Bernardins

Sciences et Avenir: How does France position itself in the development of artificial intelligence (AI) in medicine and would it be possible to imagine a database on a European or international scale?

Bernard Nordlinger : France is positioned rather well in the field of artificial intelligence, it was the effort desired by the Villani commission in 2018 and by the State, which allocated funds to develop it. We are probably not the first country in this area, but we have, for example, organized a symposium with the Massachusetts Institute of Technology (MIT), which considers us to be good partners and wishes to renew the meeting this year. In health, we are fortunate to have available a database from the vital card, accessible to researchers, subject of course to respect for privacy. This type of access to health data is unique in the world, others envy us. During the Covid-19 pandemic – a chapter of the book is devoted to it, artificial intelligence did not have many applications, which was not the case with digital technology, which was used to create programs to predict the epidemiological evolution. This is what has enabled the health authorities to anticipate influxes in intensive care, the degree of barrier gestures to be adopted, or even to manage the results of vaccination. I think this crisis has been well managed in France compared to other countries. All the information collected will be very important if unfortunately, but it is likely, we are facing another health crisis of this magnitude. We know that solid databases are the foundation for the development of artificial intelligence.

There is a body to try to manage health data at European level, but it is particularly at risk. In France, these data are collected by the National Health Data System (SNDS) from Social Security, thanks to the vital card, but also through hospitals and death registers. Increasingly, they are also stored in the Health data hub, an organization responsible for representing France at European level. To gain access, certain conditions must be met. The data must be anonymized, with the assurance of not being able to re-identify the person. They must also be used wisely and not be used, for example, by insurance companies to adjust premiums or contributions according to the risks, or by sending medical visitors where pharmaceutical companies believe that there is a marketing objective. It is obvious that this development must be adapted at European level, but certain regulations make international collaboration very difficult, particularly with the United States, which handicaps research laboratories. However, the Covid-19 epidemic did not stop at the borders of France or Europe, and to create effective vaccines required international cooperation.

Sciences et Avenir: Are certain medical disciplines more affected than others by the arrival of artificial intelligence and what are the major challenges in putting it into practice with medical personnel?

Bernard Nordlinger : These applications are gradually developing and have not really come into practice at present. Some estimate 10 or 15 years from when the use of artificial intelligence will be effective. There are certainly areas where the contributions of digital and AI are more obvious, others where it is less so, but all medical disciplines will be concerned in the end. Artificial intelligence can quickly manage huge amounts of data much more efficiently than a human, provided we know how to use it. The first area where it will provide considerable help is in image processing, in radiology of course, but also in dermatology, pathological anatomy and even ophthalmology. Other disciplines less known to the general public are also concerned, such as psychiatry, where artificial intelligence is beginning to take an important place. The machines analyze the images in a much finer way than radiologists can do, where sometimes certain elements of a few pixels are too small to be clearly identified with the naked eye.

But, a priori, and this was recalled by Cédric Villani during the conference, the obstacles to the progress of artificial intelligence in medicine are more human than technical. Physicians will have to accept training and consider AI as an aid in their daily practice, to support them in their profession by having a large amount of information in hand, without fear that their work will be taken away from them. This is particularly true for radiologists, who will have to work closely with artificial intelligence and will have to train in this tool. This training will also concern general practitioners. In China, in some places where there are not enough doctors, there are applications that can direct patients according to the severity of symptoms. In France, we will see this type of diagnostic assistance application appear, with of course a human being who validates or not this opinion. But the goal is not to go towards the replacement of the doctor by an algorithm in which one would enter his symptoms to have a diagnosis. When we are sick, we need a human being to explain a decision and inform us, especially if it is a serious illness.

There have been attempts at applications on the theme of cancer, with the program Watson Health from the firm IBM, but which did not work very well. The goal was to help diagnose cancers and several American and French hospitals had bought these algorithms, but it turned out that it was not very efficient, in particular because the data with which the AI ​​was trained came from only one medical center. But other attempts of the type will see the light of day, the internet giants are in the movement, there is for example the American group Mayo Clinic who offered to work with Google. Health is today one of the main areas of development for AI, we will see various applications appear that will undoubtedly work more or less well. However, we recalled in the book that artificial intelligence should not be fantasized as a solution to everything. The term itself is not very appropriate and the result is that it is scary. It’s a bad name that lets people think that we’re going to steal their brains, that a computer is going to take over our decisions. Artificial intelligence is not there to dispossess humans of their intelligence, but to make us more efficient.

Sciences et Avenir: With the many questions posed by artificial intelligence from an ethical point of view, how can we guarantee a relationship of trust with artificial intelligence and with the companies that develop it, and ensure that the data is not not manipulated for bad purposes?

Bernard Nordlinger : We can imagine a situation where a general practitioner, who works with an algorithm, agrees with the diagnosis proposed by the machine. In this case, the doctor explains to the patient why he agrees and validates the AI’s proposal, which implies that the doctor is trained to have an idea of ​​how it works. But the question people ask is, “What happens if the doctor doesn’t agree with the algorithm?” For the moment that has not happened, but these questions will have to be raised gradually. Generally speaking, when a doctor uses an artificial intelligence, it is he who is responsible if there are medico-legal problems that ensue, except in the case of a characterized fault in the device.

Hospitals are also prime targets for cyberattacks, which will hold health data hostage by demanding a ransom or steal it to sell it to private laboratories. There is also a risk, more insidious, of manipulation of the brain by the data or brain hacking, in order to lead people to modify their judgment by valuing certain data at the expense of others. The example that many people know is when we are asked to accept cookies when we arrive on a website. The “accept cookies” button is much better highlighted than the “continue without accepting”, which guides a decision. If you want to be a “citizen of new technologies”, you have to be educated about what they are, know the risks and know what to get out of them. It is therefore important to explain these issues, to gain the trust of health professionals but also of the citizens who benefit from these AIs.

Unfortunately, the media’s treatment of artificial intelligence sometimes inspires fear rather than providing objective information. In the book, we tried to be reasonable and to call on the best French experts in their field, to show where the research is, what the risks are and how to make good use of them. Our objective is to explain and inform in order to gain people’s trust. I asked all the authors that the texts be both flawless for specialists, but also understandable in their main principles by the public. We don’t ask people to read in one go, but those who are interested in the topic of trust, for example, can read sociologist Gérald Bronner’s chapter on trust and the feeling of dispossession by AI. There are syntheses on imaging or cancer. Once again, the goal is to explain that artificial intelligence is a tool and that you have to know how to use it.

We wish to give thanks to the writer of this short article for this awesome web content

INTERVIEW. “The obstacles to AI are more human than technical” – Sciences et Avenir

We have our social media profiles here , as well as additional related pages here.