On the 19th of June, EIF hosted a lunchtime debate in the European Parliament with the title 'Artificial Intelligence and Ethics' chaired by Emilian Pavel MEP and EIF member.
MEP and EIF member Emilian Pavel opened proceedings stressing the need to address the ethical issues raised by the prospect of ubiquitous uptake of artificial intelligence in all walks of life. This will require agreement on values and respect for human rights, ideally leading to worldwide global standards for human-centric AI. The EU must lead this effort with legislation and codes of conduct fully respecting the Charter of Fundamental Rights.
Leo Karkkainen, leader of AI at Nokia, observed that today’s digital technologies have led to a huge scaling of AI development, which will no longer be based on rules introduced by the programmer, but will be based on programming to solve problems and then analysing the available data to derive problem-solving behaviour. The ethical dimension can then be introduced by putting ethical rules on top of the data-driven behaviour. Doing so can also reveal inherent biases in the data sets or problem-solving code.
The fundamental issue for AI applications – and fundamental barrier to entry - thus becomes how to get the necessary data to drive the process. Europe is behind others, notably China with its vast population and the US with products that generate data themselves; Europe thus needs open data sets to enable widespread uptake and development of AI.
Professor Raja Chatila, Chair of the IEEE Global Initiative for Ethical Considerations in AI and Autonomous Systems, seconded this view of AI, which 1) “makes computers smarter” by giving them data (knowledge), and 2) gives them methods to solve problems. Today we see a revolution in both of these main pillars: vast capacity to crunch data and problem-solving methods developed through learning, not programmed solutions. Europeans are in the front line of these developments at national level, but lack an EU framework for further development, particularly for the ethical dimension.
The key concept driving ethical standards must be “trustworthy” AI, otherwise there will be no demand. This extends beyond economic interest notably to values and respect for human rights, with overall contribution to wellbeing the key for human-centric AI. This will require a focus on the humans behind the machines, who must be “responsible and liable for what the machines do”, which will in turn require disclosure and “auditability” to make clear the reasons for AI-based decisions. There will also need to be agreement on what decisions should not be delegated.
So how to go forward? We need more research for a strategy of “ethical by design” base on the supply of standards to industry, and independent authorities to certify compliance. This approach must also cover education at university level and workforce training, together with actual work itself.
Finally, Lucilla Sioli, Director for Digital Industry, DG CNECT, European Commission, stressed the importance which the European Commission attaches to AI development, for both its central role in future economic competitiveness and in addressing societal challenges. The Commission’s proposed AI strategy released in April recommends a wholeistic approach to mainstream AI, intended to improve European competitiveness, address ethical challenges, and anticipate social and economic impacts.
For this reason the Commission calls for €9.3 billion in the MFF, notably to support the uptake of AI at all industrial levels. For this the Commission proposes the creation of “digital innovation hubs” to support the uptake and customisation of AI at affordable cost. One focus will be to help actors acquire the necessary data.
Ethical issues will also be key, notably legal implications. We will need a set of EU level principles in order to avoid barriers within the digital single market. For this the Commission is setting up a group of high-level stakeholders to advise on these principles and how to implement them, supported by an “AI Alliance” platform through which any and all interested parties can contribute their thinking.