Responsible Artificial Intelligence

Sponsored Content

Artificial Intelligence (AI) can help us in many ways: it can perform hard, dangerous or boring work for us, can help us to save lives and cope with disasters, can entertain us and make our daily life more comfortable.

Advances in AI are occurring at high speed. The potential risks and problems of AI technology are filling newspapers (e.g. Observer, 2015, the Guardian, 2015) with discussions ranging from killer robots to privacy concerns, the consequences of AI for labour and social equality (Daily Express, 2016), or superintelligence (CNN, 2014). However, rather than being a threat to our existence or plotting to take over the rule of the world, AI is already changing our daily lives, almost entirely in ways that improve human health, safety, and productivity.

In the coming years we can expect AI systems to be used increasingly in domains such as transportation, service robots, healthcare, education, low-resource communities, public safety and security, employment and workplace, and entertainment (100 Year AI report). But these systems must be introduced in ways that build trust and understanding, and respect human and civil rights.

There is, in fact, a lot to be positive about. Currently, over a million persons die annually in traffic accidents, more than half of which are caused by human error. Even if intelligent self-driving cars do cause accidents and deaths, forecasts show a sharp decrease in road casualties associated with the increase in self-driving cars. Similarly, jobs will be lost – but maybe repetitive, monotonous, demeaning jobs should be lost, freeing up people for more meaningful and joyful occupations.

AI developments will contribute to a much-needed redefinition of fundamental human values, including our current understanding of work, wealth and responsibility – all of which will be part of the debate in the panel session AI: is the future finally here? at ITU Telecom World 2016 in Bangkok this November.

Work: As AI systems replace people in many traditional jobs, we must rethink the meaning of work. Jobs change, but more importantly, the character of jobs will change. Meaningful occupations are those that contribute to the welfare of society, self-fulfilment and the advancement of mankind. These do not necessarily equate with current ‘paid jobs’. AI systems can free us up for these occupations, allow us to be rewarded for them, to care for each other, engage in arts, hobbies and sports,  enjoy nature, meditate –  all those things that give us energy and make us happy.

Wealth: Technological developments in the last century led to mass production and mass consumption. Until very recently, having has been the main goal, and competition the main drive: “I am what I have”. Digital developments, including AI, favour openness over competition: open data, open source, open access, and so on. The drive is now quickly shifting to sharing: “I am what I share”. Combined with the changing role of work, this novel view on wealth requires a new view on economics and finance.

Responsibility: As AI moves from a tool to a teammate, perhaps the most important result of AI advances is the need to rethink responsibility. Developments in autonomy and machine learning are rapidly enabling AI systems to decide and act without direct human control. Greater autonomy must come with greater responsibility, even when the notions of machine autonomy and responsibility are necessarily different from those that apply to people. Machines are already making decisions. We need to deal with longer chains of responsibility, and with responsibility being extended to refer to machines and corporations.

Responsibility contributes to trust and includes accountability, i.e. being able to explain and justify decisions. Our trust in other people is partly based on our ability to understand their ways of doing (by putting ourselves in their place), but this does not hold true for machines. Trust in machines must then be based on transparency. Algorithm development has so far been led by the goal of improving performance, leading to opaque black boxes. Putting human values at the core of AI systems calls for a mind-shift of researchers and developers towards the goal of improving transparency rather than performance, which will lead to novel and exciting algorithms, turning deep learning into valuable learning.

Several initiatives are currently focusing on the ethical and societal aspects of AI development, including the IEEE Initiative on the Ethics of Autonomous Systems and the Partnership on AI.

I foresee an exciting future coming forth from AI developments. We are ultimately responsible. As researchers and developers, to take fundamental human values as the basis of our design and implementation decisions. And as users and owners of AI systems, to ensure a continuous chain of responsibility and trust encompassing the acts and decisions of the systems as these learn and adapt to our society.

Posted by Virginia Dignum

Virginia Dignum is Associate Professor on Social Artificial Intelligence at the Faculty of Technology Policy and Management at TU Delft. She holds a PhD from the Utrecht University, in 2014. Previously, she worked for more than 12 years in business consultancy and system development in the areas of expert systems and knowledge management. Her research focuses on value-sensitive design of intelligent systems and multi-agent organisations, in particular on the formalisation of ethical and normative behaviours and of social interactions.. She is secretary of the International Foundation for Autonomous Agents and Multi-agent Systems (IFAAMAS), member of the Executive Committee of the IEEE Initiative on Ethics of Autonomous Systems, and co-chair of ECAI2016, the European Conference on Artificial Intelligence. m.v.dignum@tudelft.nl