Cathy Li is the World Economic Forum’s Head, AI, Data and Metaverse and Deputy Head of Center for Fourth Industrial Revolution (C4IR). She is currently spearheading the Forum’s AI Governance Alliance, a multi-stakeholder initiative that unites industry leaders, governments, academic institutions, and civil society organizations, to champion responsible global design and release of transparent and inclusive AI systems.
Before joining the Forum in 2018, Li was a Partner at GroupM, WPP’s media investment arm and held various leadership positions across the vast portfolio of the WPP’s business including digital media buying, data and analytics, content, strategic consulting and financial communications in London, New York and Beijing. Li, who earned an MBA from Columbia Business School, recently spoke to The Innovator about governance in the age of GenAI.
Q: What kind of work is the Forum doing around AI?
CL: I’m responsible for the digital technology portfolio which consists of AI, data economy and policy, digital safety, digital trust, metaverse, and now focusing on digital public infrastructure, so everything about digital technology, from content to protocol.
For AI, we launched the AI Governance Alliance in 2023 that’s now the umbrella for the Forum’s AI work. Even though the name is AI Governance Alliance, it’s not just about governance. Governance is a goal but at the same time we need to understand the technology, where it’s heading and how industries are applying it. It’s important that everyone has a baseline understanding of what this technology is and how we can best leverage it for the for the maximum benefit of society. Many organizations are already applying GenAI, so encouraging responsible AI adoption now is key.
The AI Governance Alliance has tracks on technology, industry, transformation and governance. Recently we also launched a work stream on inclusive AI for growth and development which encourages the policy makers to identify and prioritize the core components and elements that are going to be key to their AI strategies from data availability and quality to access to compute, AI models, applications, talent, innovation, industry and governance. The inclusive AI workstream aims to support different regions and encourage governments to revisit their national AI strategy and evaluate them according to these eight core components and see how they’re doing, how they’re keeping track with their growth goals and where we can support them. The idea is to make sure that no region or country is left behind during this AI revolution.
Q: Who belongs to the AI Governance Alliance?
CL: Over 390 organizations, including large established companies, leading AI startups, such as Open AI, Inflection AI, Cohore, Databricks and many, many others because it is key to have the innovators working with us in this community. We’ve invited over 30 governments to join the work, as well as civil society and academia. We have just published a white paper entitled ‘Governance in the Age of Generative AI: A 360 Approach for Resilient Policy and Regulation.’ This report alone involved over 110 working group members from 27 countries across all the regions and drew from 700+ written feedback comment, twenty one-on-one interviews and six community meetings. The most important thing is to for the community to be able to drive consensus throughout this process.
Right now, we are exploring horizontal issues involving GenAI in advanced manufacturing, financial services, energy, media, entertainment, consumer and telecom sectors, among others.
The idea is to synthesize what’s happening around the world and provide practical solutions.
Q: From the work that you’ve done so far, what do you see as the major challenges posed by GenAI?
CL: There are many because GenAI is quite new, and it is quite different from classic AI, where you know the input and you can predict the output. We don’t really know how the black box inside the large language models works. GenAI introduces challenges in areas such as interpretability and traceability as you cannot easily trace it back to the input data. Accountability is also an issue. Who is accountable for what and how do you divide up the responsibility between the model producers and the downstream application developers? There are also concerns about GenAI’s impact on human interactions and society at large.
A complicating factor is that GenAI is advancing at a pace that outstrips current regulatory frameworks. We’ve dedicated several chapters in the white paper to see if GenAI regulations can be reconciled with existing digital regulations such as the ePrivacy Directive {an EU directive on data protection and privacy], the EU’s Digital Market Act, and other legislation.
There are also stakeholder specific challenges. Governments, industries, civil society and academia all face very distinct challenges in governing GenAI so tailored approaches are definitely needed for each.
Another challenge is the need for international cooperation because GenAI is a consumer product that’s taking society by storm. We don’t want a situation where there are pockets of robust regulation while other countries or regions are a sort of Wild West. Inconsistent regulations across countries can hinder effective GenAI governance. Coordinated international efforts are necessary to ensure a harmonized approach that addresses these challenges globally.
Lastly, I would say future preparedness and agility are key because alongside of the rapid pace of GenAI other emerging technologies such as quantum computing and synthetic bio are being accelerated exponentially so there’s a critical need for governments to embed agility in their governance frameworks by developing foresight mechanisms, conducting strategic horizon scanning and preparing for multiple possible futures.
.
Q: What are some of the main concerns expressed by industry and what should be done to ensure that industry has a voice in future regulations?
CL: There are significant concerns about how liability for GenAI systems will be managed under future regulatory frameworks and about what kind of data can be used versus not. Employees are already using all the different available GenAI consumer products, introducing risks, so many organizations have to react very, very quickly by putting relevant processes and governance committees in place that touch on many parts of the organization from human resources to legal to strategy and operations. Regulatory uncertainty, compliance burdens and issues related to intellectual property and liability are all of concern to industry. Together with privacy, I would say the IP issue has been the most contentious. Different countries have different perspectives based on their respective industries, how strong the different industries are, and their histories of handling IP issues. Industry leaders are concerned that overly strict or fragmented regulations could increase compliance costs and slow down the adoption of new technologies, ultimately affecting their competitiveness.
To ensure that the industry has a voice in shaping future regulations, it’s essential to engage industry leaders in the ongoing multi stakeholder dialogues. We need to engage not just the tech companies, but really everyone who’s using AI and that’s literally everyone these days. Insights and concerns need to be heard. Governments should promote public/ private collaboration by creating formal channels for corporations so that they can work together to co-develop flexible and adaptive regulatory frameworks. Fostering continuous knowledge sharing between industry, regulators and other stakeholders will help ensure that regulations remain responsive to technological advancements and align with industry needs. Governments should also invest in interdisciplinary research to better understand the ethical, social and technological implications of technology.
Q: What in your opinion is the best way to upscale government officials so that they understand these issues and they can keep pace? Foresight sounds great, but most regulators have no background in tech and have trouble keeping up with their daily agenda, never mind thinking about what will happen five or ten years from now. How do we change that?
CL: It will not be easy. We could say government should invest in targeted training programs but even just to evaluate what training programs are needed and who are the real experts is not a simple task. There is also talk about how there should be a certification for AI experts and how to scale them but there’s a real shortage of not only AI experts, but most importantly people who understand your business, your strategy, your country’s strategy, and who can really help you to implement it. That’s very different from having access to a brilliant AI researcher, because that person might not be able to help you address your specific questions. That’s why we are encouraging government officials and business executives to join the Forum’s AI Governance Alliance. Think of it as a shortcut. Whenever new technology like this comes along, you typically will have a 6-month to 12-month information black box. The best way to keep up is to join discussions within the Alliance as early as possible and to share the information with peers. We really need everyone to come together so that the knowledge sharing is much, much more effective and efficient. Along with horizon scanning and foresight a clear guidance framework should be developed and shared to ensure that government officials are aligned with best practices and can respond effectively to the evolving nature of GenAI.
Q: Can any corporate join the AI Governance Alliance or do you need to be a member of the Forum?
CL: To be part of the alliance you need to be a member. The Forum now has over 900 partners, so it is quite a big network. That said, we offer online webinars for our digital members, and we basically tried to really democratize the knowledge as much as possible. If you are interested in learning more get in touch.
This article is content that would normally only be available to subscribers. Sign up for a four-week free trial to see what you have been missing.