High-Level Summary of the Privacy and Generative AI Symposium

Held December 7, 2023 in Ottawa, Canada

---

Opening Remarks – Philippe Dufresne, Privacy Commissioner of Canada

Commissioner Dufresne highlighted the importance of gathering key stakeholders together to examine the challenges and opportunities with generative AI technology. His speech compared the current state of generative AI to the dawn of commercial aviation, when safety checks were put in place before having passengers benefit fully from the new opportunities air travel offered. The Commissioner looked at recent developments in Canada and internationally aimed at regulating AI and promoting its safe and responsible development, notably the June statement of the Roundtable of G7 Data Protection and Privacy Authorities, the October Global Privacy Assembly Resolution, and November’s Bletchley Declaration to which Canada is signatory. Adding to this the Commissioner launched principles for the responsible use and development of generative AI drafted by Canada’s federal, provincial and territorial privacy authorities. 

Keynote Speech – Dr. Gary Marcus

Dr. Gary Marcus discussed the myths and realities on the current state of generative AI. He noted that the current AI models are far from perfect, and that true generative AI is still years away. To illustrate this point, Dr. Marcus pointed to examples of publicly available large language models generating misinformation, hallucinations, or falsehoods, while appearing fluent and legitimate. Dr. Marcus also revealed that newer generative AI models are not necessarily solving these issues either, arguing that technology companies are running out of available training data while not seeing performance improvements. Finally, Dr. Marcus argued that the interests of technologists do not align with the interests of humanity, and that effective AI regulation requires participation from governments and independent researchers.

Panel 1 – Generative AI as a Technology: Opportunities and Risks

This session underscored that AI governance is a collective responsibility of humanity, and that while it can provide incredible benefits, there must be effective regulation to prevent negative outcomes. Panellists from industry, government and academia spoke about how advancements in real-time processing power have enabled novel uses for AI in the healthcare and automotive industries. They also remarked that AI has the power to spread disinformation, undermine traditional labour markets, and worsen bias and racism towards minorities and racialized communities. Panellists called for greater international coordination in the terms and governance given to AI technologies, with a view to enhancing citizens’ trust.

Panel 2 – Generative AI Governance, Standards and Regulation

Panel 2 focused on the role of standards, international agreements, and domestic legislation to effectively regulate generative AI technologies. They pointed to the Artificial Intelligence and Data Act (part of Bill C-27) as providing a high-level framework for AI governance that can provide flexibility without limiting sector-specific use and regulation. Panellists spoke of the role of government in implementing these standards and promoting a multi-faceted approach to AI regulation that includes broader understanding of use cases. Further, they noted that while commonly accepted definitions and standards will facilitate global interoperability, these need to be aimed at a greatest common standard, not a least common denominator – and getting to that will take time and continued effort.

Panel 3 – Innovation and Human Rights in the Context of Generative AI

The Symposium’s final panel looked at the relationship between generative AI technologies and human rights. Panellists discussed how AI technologies can be used to infringe on human rights including through surveillance, automation bias and misinformation. Panellists spoke of the need for more transparency when developing AI systems, such as by having civil society, academics and the communities most directly impacted by the technologies be more directly involved in their development and deployment. They encouraged the development of regulatory frameworks that instead of being complaints-based, are proactive and focus on the obligations and responsibilities of the different actors involved.