Principles for AI: Towards a humanistic approach?

20 Mar 2019 01:00h

Event report

The conference was organised by UNESCO, in collaboration with the government of Japan. The first part of the event was featured by introductory speeches framing artificial intelligence (AI) as a phenomenon creating important opportunities and challenges to modern society. Co-operation among different stakeholders has been reiterated as crucial in the definition of principles to apply in the development of AI. The conference was then featured by two panel discussions focusing on the challenges and opportunities of AI, and on the approaches needed to achieve universality in AI governance.

The conference was introduced by Ms Audrey Azoulay (Director-General, UNESCO) who underlined the unique feature of the technological revolution. She stressed that we must be aware of the shared responsibilities involved in tackling such a revolution and put AI at the service of the common good. Recalling Asimov’s lessons, she explained that robots fascinate us because they question who we are, what we do, and what we are capable of. Following this pattern, she argued that AI questions our own humanity: e know that certain categories of jobs will disappear and that others will be left behind. As a result, she stressed UNESCO’s mandate in sharing knowledge and expertise about the application of AI for the common good from developed to developing countries. Finally, she concluded by underlining that AI is defined by the data provided: data represents the choices we have made. In this context, all data-related issues have to be discussed before they affect society and they are used by society. This needs to be discussed on a global scale with a multistakeholder approach and not just among technical scientists.

Additional remarks were addressed by Mr Angel Gurria (Secretary-General, OECD) who talked about how to bring human values to the heart of AI. While AI is driving optimism, it is also fuelling concerns such as the dangers of codifying and reinforcing existing biases infringing human rights and values. There is a need to design more transparent and accountable AI systems. This represents a challenge that no single category can address by itself: as a result, there is a need for a global multistakeholder response to a global issue. In 2018, the OECD established a group of experts to scope a set of principles to facilitate innovation and the adoption of – and trust in – AI, based on the following principles: human-centric values, fairness, transparency, robustness, and accountability of AI systems. Moreover, the OECD is establishing an AI Observatory Policy Lab with the goal of ensuring consistency and complementarity between the OECD’s work and other international activities.

Mr Tapiwa Chiwewe (Research Manager, IBM Research Africa) talked about AI aiming to provide a technical perspective while demystifying AI. He argued that AI represents the ‘new IT’. As information technology (IT) was a key enabler of the information age, AI is enabling the generation of new information from numerous different sources in a new ‘cognitive age’. AI is about performing cognitive functions benefiting from the immense amount of data currently available, which is meant to exponentially grow in the following years. The rapid growth of AI has been facilitated by variables such as the proliferation of data and the advance in algorithmic software, which have be translated into machine learning, neural network, and deep learning. After the that technical introduction, Chiwewe explained that the principle at the basis of IBM’s research on AI is to augment and not replace human intelligence. This can be only done if there is trust in the technology and it is based on the principles of fairness, explainabiliy, robustness, and accountability.

Dr Cédric Villani (Member of the French Parliament, AI expert and 2010 Fields Winner) talked about the pathway that brought AI to the core of international discussions. In the 1950s, the goal was to understand human intelligence in order to reproduce it. Today, the goal is to reproduce not just the intelligence, but the actions. Algorithms have improved to the extent that AI driven processes can win in strategy games such as chess or Go. However, AI is still in a state of experimental development which needs further research to understand which fields and applications can be most successful. Therefore, it is crucial from an institutional and policy makers’ perspective to facilitate experimentation and to organise the research. Moreover, another important aspect is the notion of sharing experience, data, results, and knowledge. Finally, the sovereignty of human beings is critical for the development of AI.

The first panel discussion was moderated by Ms Cathy Mulligan (Visiting Research Fellow, Centre for Cryptocurrency, Imperial College, United Kingdom) and addressed the opportunities and challenges of AI.

Mr Omar Bin Sultan Al Olama (Minister of State for Artificial Intelligence, United Arab Emirates (UAE)) argued that the biggest challenge is inclusion. Furthermore, he stated that rather than being absolutely negative or positive about AI, there is a need to shape its evolution according to agreed upon principles. He gave an overview of the UAE’s approach to AI. First, efforts were focused on putting in place a system able to attract experts pushing research. Second, the government is deploying AI in key sectors such as healthcare, tourism, safety, and security. Finally, the UAE plans to become an exporter of AI expertise rather than an importer of it. To finish, he argued that in order to discuss the ethical considerations of AI, a better understanding of the use cases of AI needs to be achieved.

Mr Seng Yee Lau (Senior Executive Vice President, Tencent) stressed the ethical considerations to keep in mind when discussing and shaping AI. Furthermore, he stressed the importance of having an open ecosystem able to promote the full potential of AI. In addition to that, it is important to have an active will of the private sector to complement the policy makers’ activities and sharing responsibilities in addressing the raising challenges of AI.

Ms Bunmi Banjo (Managing Director, Kuvora Inc.) argued that we have not seen policies being made at the same speed as technological change. She stressed the crucial role of education from a research perspective as well as from an individuals’ perspective. In addition to that, she underlined the importance of a public-private collaboration to maximise the benefits of AI and to address its challenges.

Ms Marija Manojlovic (Strategy, Data and Innovation Advisor, Global Partnership to End Violence Against Children) stressed that violence and discrimination is perpetrated in data and therefore in the applications of AI. She further explained that algorithms are increasingly used to make important court decisions; however, discrimination is still perpetuated in the data used which represents one of the challenges that policy-makers need to address

Mr Nicolas Economou (Founder and CEO, H5) argued that if we want to maximise the benefits of AI for sustainable development, we need to be institutionalist in thinking about AI governance. He stated that in certain domains and applications of the law, AI can advance the function and values that animate the law. Nonetheless, it is extremely hard to discuss AI governance in abstract. In addition to that, he strongly stated that the field of the law has the advantage of knowing and already possessing a set of agreed upon and shared principles: the task is to translate them to the application of AI. Finally, he recalled the work done by the IEEE with the Global Initiative on Ethics of Autonomous and Intelligent Systems in enhancing principles and ethical considerations in AI.

Mr Davor Orlic (Chief Operating Officer, Knowledge 4 All Foundation) argued that in order to bridge the divide between AI and governance, the first step is to understand the context and contextualise challenges and opportunities.

The second panel discussion was moderated by Mr Andrés Roemer (UNESCO Goodwill Ambassador for Social Change and the Free Flow of Knowledge; Writer; Philanthropist; Human Rights Activist) and addressed the notion and conceptualisation of a fully inclusive approach for AI.

Prof. Abdoulaye Baniré Diallo (Professor, Université du Québec à Montréal, Chief scientist and Co-founder, My Intelligent Machine, Canada) stressed the importance of guaranteeing equal access, regardless of the whether the areas of provenance are developed or developing. The gap between developed and developing countries is still big: efforts are needed to tackle this divide by supporting developing countries to access educations in order to enable development.

Prof. Raja Chatila (Professor of Artificial Intelligence, Robotics and Ethics, Sorbonne University, French Republic) talked about the work of the IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems in enhancing principles and ethics while providing a ground for establishing standards. These standards have ethics at their core, and highlight the need to provide ethical by design AI developments.

Ms Anriette Esterhuysen (Senior Advisor on Internet Governance, policy advocacy and strategic planning, Association for Progressive Communications (APC)) argued that opportunities and challenges need to be balanced and complemented by deeper education to address governance challenges to achieve inclusion. Furthermore, she reiterated the importance to include human rights considerations when addressing AI challenges.

Mr Miguel Luengo-Oroz (Chief Data Scientist, UN Global Pulse) argued that thinking about universality means thinking about long terms scenarios in which solidarity should be the principle of AI. To complement the picture, he stressed how crucial it is for AI experts to be complemented by experts in other fields from around the world to achieve the goal of universality.

Mr Nicolas Miailhe, (Co-Founder and President, The Future Society) argued that AI has to be nested in the wide global governance revolution and that the keywords to address the issue are pluralism and social contract.

The panel discussion then focused on ensuring a multi-stakeholder approach.

Mr Ulrik Vestergaard Knudsen (Deputy-Secretary-General, OECD) recalled that AI is going to transform every aspect of our lives, raising challenges that need to be addressed through a multi-stakeholder approach. He argued that other than engaging engineers and scientists, philosophers and psychologists need to be included to complement the picture.

Ms Hawa Ba (Head of the Senegal Country Office, OSIWA) stressed that the inequalities we are witnessing in the world are due to the biases present in AI practices. Universality can be achieved if inclusivity is featured in the application of AI at all levels. She further argued that the challenges posed by AI need to be addressed both locally and globally.

Mr Katsumi Emura (Chief Technology Officer, NEC) explained how by understanding the goal of achieving universality in the governance of AI, it is necessary to communicate and engage with every stakeholder to design the most inclusive and fair ecosystem. Finally, he acknowledged that technology and AI are neutral, it is their design that is featured by ethical considerations and implications.

Mr Jernej Pikalo (Minister of Education, Science and Sport, Republic of Slovenia) explained that AI is an enabler of perceiving different views of our society. He argued that to be able to regulate, however, there is a need to understand all the issues that come up around the application of the technology itself. This is the mandate of the Center for Artificial Intelligence that Slovenia is establishing as part of a bigger network of national and international labs.

Mr Francois Taddei (Director of the Center for Interdisciplinary Research, French Republic) argued that reaching inclusivity means engaging with different stakeholders by inventing new means. There is a need to prototype experiments and develop collective intelligence.