Last month, the European Commission published a White Paper outlining the European Union’s approach towards creating an environment of excellence and trust in Artificial Intelligence technology.
The Commission’s overarching goal is to unify policy across Europe so as to both prevent fragmentation in the AI tech space, and to ensure that European values and ideals are implemented in future AI applications.
The key points I will be discussing relate to the ethical considerations of AI as addressed within the paper, the categorical definitions proposed of ‘high-risk’ vs. ‘low-risk’ AI, and the implications of rapid financial investment. Subsequently, I will highlight the especially crucial aspects which must not be left behind in the ever-present global technological race and implore the EU to act on its legislative precedent by putting the rights of consumers before those of industry.
To begin, the two pronged approach is guided by the concepts of excellence and trust. In order to achieve excellence, the Commission proposes increased investment in adoption of AI tools across industry. As for trust, their focus is on consumers by building Human-Centric AI which are guided by the Ethical Principles agreed upon by the High-Level Expert Group on AI as informed by the OECD’s framework.
The White Paper also further distinguishes between two general categories of AI applications, those deemed either ‘high-risk’ or ‘low-risk’ by the Commission, with differing obligations enforced upon creators of AI tools depending on which category they serve.
‘High-risk’ AI constitutes organizations which employ AI in situations where risk is likely to occur, such as healthcare, transport, and energy to name a few. Risk will be judged according to impact on affected parties where the legal rights of individuals or companies may suffer or where injury or death can occur. Unsurprisingly, high-risk AI will be subject to more stringent regulation in the realm of data storage, information provision, human oversight and accuracy.
However, the regulation of ‘low-risk’ AI technology is of equal importance. For this, there is a call to follow the regulations put forth within the paper in a self-regulatory way. Those who participate will be awarded a quality label in an attempt to create the desired bond of trust between AI and humans, especially within the consumer, nonprofit and public service sectors. We as an organisation would look forward to participating in a scheme such as this so that concepts such as ethics by design, which we already incorporate into our products, can have a more widespread adoption.
While some criticism has already been levelled at the difficulty of defining the two types of AI, it is important to remember that this is still a White Paper. That being said, furthering the discussion around the new policy proposals is a crucial part of policy creation, so I would encourage any thoughts and comments down below. I should probably mention that I will be writing a consultation response to the White Paper, but more on that later.
We at AI for Good are pleased to see that regulatory policy in the AI space is considering the ethical implications of the technology, and see the positive benefits which can come about by ensuring that the development of AI throughout various European countries is informed by the same foundational principles.
Some of these benefits are the ones discussed by the paper itself regarding the unification of consumer liability protection in AI products and a continuing understanding of the mental safety risks of users. Beyond this, I would argue that such attempts at creating comprehensive regulatory frameworks, will prove influential on a global scale and hopefully provide the necessary guidance towards future developments in AI technology beyond the European space.
In terms of the scope and ambitious regulatory undertaking which this White Paper is introducing in the AI sector, I believe that there can be a lot of parallels drawn with the related Department for Digital, Culture, Media & Sport’s Online Harms White Paper regarding digital platforms, especially in the areas of harms and risks to users. Moreover, as the EC’s White Paper acknowledges, there will be a need to update existing concepts regarding online consumer safety as well as product safety legislation, with liability of post-market entry AI integration considerations being particularly important. To help address these issues, I think that the EC would benefit from adopting a ‘duty of care’ approach like the one behind the UK’s Online Harms proposal in so far as it seems to provide a useful conceptual way of clarifying why ethics by design are important in AI development not just in a legal sense, but in a societal one as well, as digital spaces continue to become larger places of public social interaction which physical spaces once were.
That being said, if the duty of care approach is co-opted by the EC in a future version of proposed regulation, it must be careful regarding the cost of compliance in order to avoid causing a ‘chilling effect’ surrounding general AI adoption, a criticism already leveled at the Online Harms proposal and the ramifications for digital social media platforms if they cannot afford to comply.
The Commission also calls for immediate investment to the tune of 100 million Euros in Q1 of 2020 towards the financing for innovative developments in AI provided by the European Investment Fund. The long-term plan proposes the development of a curriculum for AI developers which should be made available as a resource for training institutions by transforming the assessment list of ethical guidelines. For now, this seems to be specifically targeted towards universities and higher education institutions to attract global talent and develop lucrative AI masters programmes.
Lastly, there’s a push towards creating a dialogue between the healthcare, rural administrator, and public service sector, along with promises to facilitate greater procurement of AI systems through the establishment of an ‘Adopt AI programme’. The latter is intended to increase the number of AI tools these sectors will have access to.
In my view, the development of such an international policy approach will serve as a much needed foundation in creating an open dialogue between numerous stakeholders. Just like GDPR regulations, which have influenced legislation across the world in California with the similar CCPA laws, the enforcement of these proposals would expand wider than the immediate European community as international companies will be forced to comply with them and therefore hopefully adjust their AI policies throughout their organizations.
I believe that the European Commission is setting a good precedent in its attempt to establish a more universal, albeit within the confines of Europe, approach towards our understanding of how machine-human interactions will continue to evolve in the near future. That being said, the language which permeates throughout the White Paper is one which continually addresses the concern of the European Union falling behind other countries when it comes to establishing itself as an important player in the AI and high performance computing world. This is most likely fueled by headlines such as: ‘Europe is no longer an innovation leader. Here’s how it can get ahead’ where the World Economic Forum laid out a suspiciously similar strategy last year to the one found in the White Paper, suggesting addressing three key issues: Funding, Fragmentation, and Frame of Mind. I would not be surprised if many of the suggestions were used to develop the strategy behind the White Paper’s proposals.
With this underlying sentiment, it is my concern that despite the sections designated towards ethical considerations and framework developments mentioned within the paper, these sorts of important aspects of technological growth will be left behind as Europe comes to view AI advancements through the perspective of industry and chooses to put a greater importance on Europe’s competitive position first over its consideration of ethics in AI creation. There is a similar argument currently being made by 5G proponents, who argue that industry applications of the technology will be the economic drivers behind rapid infrastructure development in Europe, leading to the necessity of ‘network slicing’ and the abolishment of net neutrality that comes with this. Despite these arguments being in clear opposition to existing regulation protecting net neutrality, it is obvious that a wider repudiation of a fundamental principle of internet access, namely that all internet traffic should be treated equally by ISPs, is at risk of being lost forever if regulators decide to trade a contemporary infrastructural advantage to benefit industry, for the long-term digital well-being of consumers.
As the pileup of headlines surrounding Europe’s lagging behind in the global tech race call for further investment in R&D, along with similar risk capital strategies to the growth-oriented tech-company mindest found in US unicorn companies (read more on this strategy’s issues here), I fear that a similar view to Silicon Valley’s Facebook à-la ‘move fast and break things’ may be viewed as the only way to catch up to tech giants in the US and China. While competition and R&D investment are not inherently bad, I hope that the European Commission will continue to follow its precedent set by legislation such as the GDPR and the EU e-Commerce Directive, which establish influential regulatory and accountability metrics that protect consumers, and will prioritise the ethical considerations that emergent AI applications will have on our societies.
The White Paper provided by the Commission is still subject to change as consultation responses continue to be accepted until May, 31st. As mentioned above, I will be writing a consultation response, so I would love to hear any feedback and thoughts on how the policy proposals can be improved, some of which I will try to incorporate within my response!
That being said, we would encourage people, business and civil society organisations to participate as well in order to stimulate the debate on how further AI regulatory policy should take shape in the ongoing consultation. Those who wish to learn more about the White Paper you may find it here. To look at a more detailed bullet-point summary provided by me, click here.