Transatlantic reflections on AI

Artificial Intelligence, more commonly known as “AI”, often conjures images of blockbuster movies and sci-fi thrillers. But it is no longer a theoretical, future development — AI is already here and all around us.

At RELX, we have a record in developing AI applications: from protecting consumers against fraud when they access Netflix, to diagnostics to keep you safe on a Boeing aircraft. It was our AI recommender technology that enabled the BioNTech scientists to swiftly refocus their mRNA technology for the purpose of combatting Covid-19. RELX has been at the forefront of developing AI technology with commercial applications for some time.

We are also a company committed to the responsible use of this emergent technology. As public policymakers look to develop a regulatory framework for AI, we are enthusiastic about fostering dialogue about the risks and benefits AI brings to consumers and how it can transform industries. We look forward to sharing our learnings as both developers and users of AI.

That’s why last month we convened a conversation with key stakeholders on both sides of the Atlantic, which was hosted by Member of European Parliament Dragos Tudorache, Chairman of the Special Committee on Artificial Intelligence in a Digital Age in the European Parliament and Rapporteur for the EU’s AI Act.

The event was moderated by RELX’s Head of Government Affairs for the EU and global AI policy lead, Elizabeth Crossick. Speakers included Dragos Tudorache and US Congressional Representative Don Beyer, Chairman of the United States Congress’ Joint Economic Committee. Titled ‘Building a transatlantic bridge for AI and Tech’, this was the most recent event in the ‘AI Going Global’ series that MEP Tudorache has been hosting in his capacity as a leading EU politician on the topic of AI.

The event also brought together Juha Heikkil?, Adviser for Artificial Intelligence at the European Commission; Paula Goldman, Chief Ethical and Humane Use Officer, Salesforce; and Philippa (Pippa) Scarlett, Head of Global Government Affairs, RELX.

The panel explored several themes. They included what appears to the regulatory convergence between the US and the EU on AI, the importance of research-based methods to advance trustworthy AI, and how regulation keeps pace with rapidly advancing innovations. There was interest in how stakeholders can engage with legislators to ensure AI regulation is fit for purpose, avoids unintended consequences, as well as the importance of an interdisciplinary approach based on openness, inclusivity, humility, and cooperation. Importantly, as a business that is committed to the responsible use of AI in alignment with public interest, our event also explored how businesses themselves can and should implement policies to advance responsible use of AI, regardless of legislation.

Below is a more comprehensive overview of the event’s discussion points.

“SHARED FUNDAMENTALS”

A major (and somewhat surprising) theme of discussion throughout the session was how aligned the MEP and Congressman were in their thinking on regulatory approaches to AI and its applications. While there was a recognition that both jurisdictions might not reach the same ultimate solutions or frameworks for technology, our panelists agreed that a common language to describe issues and approaches was critical to successful transatlantic dialogues. In recognizing that the EU and US may ultimately pursue different regulatory frameworks, there was still appreciation that lessons that could be learned and experience shared between both sides.

As the EU Commission’s expert on AI, Juha Heikkil? pointed out, California, Maine and other US states are already implementing regulatory measures similar to the EU’s, most particularly on the use of facial recognition. Meanwhile, Congressman Beyer emphasized how it was essential AI be regulated not just by the US, but “as part of a coordinated effort” with the EU.

One way a co-ordinated, cross-Atlantic approach could achieve this is through the US-EU Trade and Technology Council. Formed in Pittsburgh in September 2021, the council will facilitate dialogue across a range of socio-economic legislative agendas and policies. In particular, the council?has recognized?that AI technology has the potential “to bring significant benefits to our citizens, societies and economies”, as well as helping to tackle significant challenges and improving quality of life.

Alongside support for the OECD’s recommendations on AI, the speakers also underlined how the EU and US are committed to working together to ensure AI is used in ways consistent with “common democratic values and human rights”. Positively, the US and EU have both adopted a risk-based approach. The European Commission via the proposed risk-based AI Act; the US via developing an AI Risk Management Framework. All panelists were optimistic that such dialogues and discussions would be beneficial in terms of future regulatory advances.

A final key insight on regulation was that both regions must be considerate of the way regulation is set up. This is to ensure that it remains adaptable in a rapidly changing and evolving market. MEP Tudorache explained that the EU was innovating in this space by moving important elements of regulatory text, such as definitions and use cases, to the annexes of legislation to allow easier “adaptability over time”. Such moves would help to ensure future legislation does not become outdated if unexpected or unforeseen advances in technology are made.

MULTI-INPUT CONVERSATIONS

An interesting thread throughout the webinar was how regulators can understand what happens “in the real world” of developing and using AI. MEP Tudorache argued the need to understand “the link between what developers are doing and what users are doing”, around algorithms that developers put together. This meant that “more than ever” regulators would “need to listen and need to understand” perspectives from companies at the forefront of AI technology.

With the perspective of having worked across business, government, and law enforcement, RELX’s Pippa Scarlett emphasized that multi-input conversations were necessary to avoid “unintended consequences” in such a dynamic space. Noting that a diverse set of experiences around the negotiating table would bring to light different questions and experiences which would help regulators and businesses to think about potential misuses, she highlighted that the business community “stands ready to partner and help work with regulators towards that end”.

WHAT ACTION BUSINESSES CAN TAKE

Engagement by governments is, of course, essential to the successful and ethical use of AI — both now and in the future. One takeaway from the session was that businesses should continue to act proactively to anticipate and avoid misuses of AI as the technology continues to evolve, RELX’s Pippa Scarlett stressed.

Paula Goldman outlined how Salesforce employs a multi-stakeholder approach to empower customers to use AI responsibly. Using an “AI model builder” — developed by the firm for use across a range of applications — the model provides insights and tools to help customers identify disparate impact in their models. Goldman added how, if a user wanted to exclude race as a variable, for example, the model builder will identify other variables in their dataset that are highly correlated with race, like zip or postal code, and would therefore act as a proxy for race. The AI model builder is an example of how AI businesses can help their customers unlock AI in ethical ways.

Alongside the Salesforce AI model builder, we heard how the creation of an internal office within the business — to set policy on how customers use their AI technology — was an approach becoming more widespread within the technology sector. The benefit of this is that it allows issues and concerns to be integrated into the processes of the engineering and design teams at the conceptualization stage, well before customers approach the firm to use their algorithmic expertise.

Another important step which businesses can take, which Pippa outlined as important to RELX, is to anchor the approach to AI in a “human centric, ethically designed way, which is respectful of human rights and driven by applications for improving lives and protecting the planet.” Sharing her learnings from government and law enforcement that legislation “can show up too slowly”, Pippa emphasized that a great opportunity existed for businesses grappling with issues to share their experience and expertise with elected officials, civil society groups and other stakeholders, as a participant in a multi-stakeholder discussion.

TRANSATLANTIC CONSENSUS ON AI

The discussion highlighted exciting developments — notably that both the US and the EU were eager to be leaders on AI, and to do so in a collaborative fashion.

As the Biden Administration brings new hopes for renewed transatlantic relations, AI has emerged as a key area for cooperation, as the recently launched TTC demonstrates. The event was therefore a timely moment to bring together policymakers and industry representatives from both sides of the Atlantic to exchange views on transatlantic cooperation on AI and tech, the key ingredients to promote it, and potential obstacles we can overcome together.

There is reason, once again, to be optimistic about the use of AI in the future.

Watch the full event here: