July 5, 2018 | Author: Patricia Morén / neuromimeTICs.org
The ability of intelligent "machines" of self-learning, following the rapid development of artificial intelligence (AI), has set off alarm bells about the urgent need to develop a universal framework of ethical principles related to AI. It is urgent to address this issue in other that the maximum number of countries can adhere, preserving their cultural differences, as well as their beliefs and values, says in an interview to neuromimeTICs Dafna Feinholz, chief of Bioethics and Ethics of Science Section (SHS/BIO) of the Human and Social Sciences section of UNESCO (the United Nations Educational, Scientific and Cultural Organization) and one of the speakers in the 2nd edition of the AI for Good Global Summit, held in Geneva (Switzerland) at the ITU headquarters from May 15th to 17th of 2018. According to Dafna Feinholz, the ethical principles that could not be renounced in an universal framework are the algorithmic transparency, the assignment of responsibility for an action of the AI and the equal access to the benefits of the AI, as well as the non-discrimination and the protection of confidentiality and autonomy.
Don’t you think that artificial intelligence is being seen more as a threat to our society than as an opportunity?
Do you really have that impression? I don’t think so.
Well, during the AI for Good Global Summit there have been some speakers who have remembered the danger posed by the development of autonomous weapons or autonomous cars guided by artificial intelligence.
In the case of AI, like in any other technological development, two potential aspects must be taken into account: that it may have benefits, but also risks. In this regard, there are two fundamental issues that we have to consider: who assesses the risks and benefits and who defines them. This question is fundamental since all decisions about what is good and what is bad are based on values. What is good and what is bad is nothing else than values. It is ethics by definition.
On the other hand, not everyone has the same idea or set of values and will not conceive the benefits of AI in the same way. This definition of benefits and risks is also part of bioethics. UNESCO has played a central role in this regard, and it is there where it has become clear the importance to involve all those affected in the development of technology and its applications and, for this purpose, collective reflection has to be carried out: to define the scope of the benefits and risks of AI and reach a common basis in which everyone involved can agree.
We have to define the scope of the benefits and risks of AI and reach a common basis in which everyone involved can agree
What work is being done by UNESCO in this regard?
UNESCO created a Bioethics Program in 1993, the result of which the only universal bioethics declarations have emerged: the Universal Declaration of the Human Genome and Human Rights; the International Declaration on Human Genetic Data, the Universal Declaration on Bioethics and Human Rights; and the Declaration of Ethical Principles in relation to climate change. But AI raises additional problems, some of them shared but others not. This additional component is that machines can learn by themselves, which poses philosophical problems: Where are the limits between the human and the non-human? When implementing an application, who is the responsible and how can we measure this responsibility?
So, will UNESCO promote work on joint documents related to AI?
UNESCO already works on AI topics in education, in communication, in the natural sciences and, of course, in the ethics of science and technology. COMEST, the World Commission on the Ethics of Science and Technology, published in 2017, a report on ethics and robotics, and is currently working on one on the ethical aspects of the internet of things. A very important and interesting topic is the relationship between AI and creativity. For example, we must ask ourselves what happens to the rights of artists? And philosophically creativity has always been conceived as an exclusively human feature, but what happens when a machine is capable of creating, of writing a story? UNESCO wants to position itself as the international multidisciplinary forum, counting on civil society, to discuss about AI and to reflect on the possibility of having an ethical framework for the development of AI based on ethical principles in both research and applications.
Why, is it because until now the documents that have been prepared are rather partial guides or come from unrepresentative forums?
There are already guides and not all are small. The European Union will present guidelines of ethical principles in AI, which is not a non-negligible forum. But this type of ethical framework in AI has to be universal. It is necessary that there is a consensus and adherence to principles in a global way and not only in a regional level.
An ethical framework in AI has to be universal. (…) it is necessary that there is a consensus and adherence to principles in a global way and not only in a regional leve
And you mean that UNESCO is more diverse.
UNESCO, by definition, is international, bringing together more than 195 countries from all continents. Let's say that you have a comprehensive view of the issues related to the development of science and technology, for its link with culture and education. The role of UNESCO is to bring to the table actors who do not see a specific issue so close or which is not so easy to involve. UNESCO has the vocation and commitment to the diversity and generates an ethical framework to develop and apply technology based on universal values, despite the differences, that are also appreciated and respected. UNESCO is in favor of cultural diversity: despite the differences there are always certain universal values that we all adhere.
As an example, for the Universal Declaration of Bioethics and Human Rights, a consultation was made to different religions and consensus could be reached.To adhere, it does not matter if one shares the same religion or not. In addition, it is important to remember that, practically, no religion is monolithic. To develop these statements there is a flexibility, they do not arise from something monolithic, but from a common basis so that different people, regardless of where they live or where they come from, can be recognized in them.
What are those universal values that could not be renounced in a future possible universal declaration or elaboration of AI ethical principles by UNESCO?
The transparency in how algorithms are programmed; the responsibility, since there must be someone responsible for the actions of the AI; and the equity in the access to the benefits of artificial intelligence, as well as the non-discrimination and the protection of confidentiality and autonomy.
The AI can serve us, but it should never be a substitute for the doctor-patient relationship
In fact, in the AI for the Good Global Summit, there has already been a little debate on this need to agree on principles in various areas, including the healthcare system, a session in which you have been one of the speakers.
In Medicine it is very important to judge the reliability of the source, when an application for phones is been build, for example, where does the information come from? It is important to be able to communicate with the patient and determine the responsibility of a medical act, despite having relied on AI. The AI can serve us, but it should never be a substitute for the doctor-patient relationship. It must be a gate to access the doctor and it is essential to educate the doctor so that he or she knows how to take care of their patients. The Hippocratic oath already said "first, do not harm". Now, despite the time constraints of the healthcare system to the doctors and that the doctors have little time, the doctor-patient relationship must be maintained.