EU Industry Day: Europe & an AI driven future
On the 22nd of February, Orange organised a workshop on Artificial Intelligence (AI) during the “EU Industry Days”. This annual event brings together stakeholders from a large number of industries and representatives of the European institutions to exchange on the Commission’s industrial policy strategy and actions the EU needs to further take to increase its industrial competitiveness.
This workshop discussed ways to balance business opportunities with ethics and responsibility in the development of AI. Divided in three parts, it first addressed the question of the definition of AI, then the opportunities for the European Industry and finally the ethical and societal issues to be tackled to gain acceptance.
Amid dreams and fears there is a need to demystify Artificial Intelligence
Nicolas Demassieux (Orange Senior VP Research & Labs) and Andrea Renda (CEPS Research Fellow) exchanged views on the concept of AI, what it can achieve today and its possible, not fully known evolution.
The debate notably addressed the challenges of machine learning where ‘machines’ become capable of learning by themselves. This evolution raises opportunities but also many questions and fears. Speakers agreed that, while we can understand how it works it is currently less easy to understand or predict all the outcomes of the different applications of algorithms. They also flagged that AI and human brains are different: AI works and “thinks” differently. Finally, at this stage of development, AI solutions while being very good at some specific tasks do not reproduce the highly diverse cognitive capabilities of human beings. What we have now is qualified as ‘weak AI’, as opposed to ‘strong AI’, and speakers recognised that we are very far from what human beings can achieve.
Asked by the moderator Jennifer Baker, speakers debated on the need to introduce regulation for AI, or to develop a precautionary principle, with some nuances; from starting now to think about regulation, having the governance right, or calling for pragmatism ensuring we focus on what is practically occurring and then assess whether there are problems to be solved. One issue mentioned related to the risk of having interconnected robots. What seemed however to gather consensus was to call for citizens to be involved from the beginning, to work on AI biases and ensure a responsible development of AI.
The clock is ticking; how to make AI an opportunity for the EU industry
Moderated by Lucilla Sioli (Director Digital Industry at DG Connect, European Commission), this panel brought together Bénédicte Javelot (Orange Chief Strategy Officer), Ondrej Socuvka (Google EU Senior Public Policy Manager), David Sadek (Thales VP Research, Technology and Innovation) and Yann Lechelle (SNIPS COO).
Lucilla Sioli introduced the debate by highlighting notably that the European Commission will adopt its strategy on AI by the 25th of April. It won’t include legislative proposals but will lay the foundations of future EU actions on this matter.
This panel focused on the steps Europe must take to be at the cutting edge of the development of AI. Most participants agreed that the EU should speed up on AI, to avoid lagging behind and also face increasingly fierce global competition. They considered that AI is a game changer, with huge potential to transform businesses and their products or services. What emerged from the debate is the need to build an ecosystem which will increase European industry’s ability to innovate. Panellists also mentioned the need for the EU to invest not only in technological development but also in human capital to develop skills.
Finally, they each delivered some recommendations to boost AI development in the EU, such as:
- Increasing trust in AI; by notably creating a trustworthy platform with tools to control or audit AI biases, developing a transparent AI, or working on certified or trusted AI by design;
- Increasing information and education to facilitate social acceptance and adoption of AI; Panellists all agreed that Europeans, both as consumers and citizens, should be involved in AI development to better understand what AI is and be aware of what AI can and cannot do;
- Re-enforcing the local ecosystem and start-up through a sort of a Small Business Act for the EU;
- Inviting companies and engineers to rely on data protection and security by design, GDPR being seen as a possible competitive advantage by several;
- Developing open AI tools to allow use of machine learning by all;
- Securing the intellectual property of AI;
- Ensuring the EU becomes not only a consumer of AI but also a producer and a leader in some sectors.
Ethical and societal questions raised by AI
Moderated by Eva Kaili (Member of the European Parliament), this panel brought together Nozha Boujemaa (INRIA Director of Research & DATAIA Institute Director), Liam Benham (IBM Europe Vice President, Government and Regulatory Affairs), and Laurent Zibell (IndustriAll European Trade Union, Policy Adviser, Industrial Policy). This panel reflected on the ethical and legal questions raised by AI development in society.
Panellists considered that it’s appropriate to say that AI developers can’t not decide alone what is ethical, or define what is socially acceptable. There was consensus that understanding, transparency and trust should be developed from start to achieve a responsible AI and gain acceptance.
The speakers considered a range of suggestions to put ethics into practice, among which:
- No need for more or new law but ensuring instead concrete implementation of existing laws under the control of algorithms to check compliancy;
- Developing a transparency and responsible or accountable AI and algorithms by design, ensuring fairness and equity in AI;
- Ensuring transparency on data ownership; on how algorithms reached decisions; on the sources of the data used and on how AI has been trained; improving data governance;
- Ensuring the information asymmetry between the AI service providers and end-users is reduced;
- Working on robustness and accuracy of big data and AI techniques;
- Examiner la problématique de l’accès aux données dans des conditions raisonnables ;
- Working on the issue of access to data in reasonable conditions;
- Guaranteeing that humans keep control over AI-based products; considering the evolution of AI, the necessity to develop a kill switch was mentioned.
Finally, the panellists also discussed the impact of AI on the labour market. Tasks will evolve and certain tasks or jobs will disappear while others will be created. They agreed that the EU, its businesses, public authorities, have to anticipate and train citizens allowing them to get the skills needed to thrive in an AI driven future.