On 13 November 2019, I participated in a panel entitled “A Fusion of Solutions: Towards Common Principles on Artificial Intelligence” with Kersti Kaljulaid, President of the Republic of Estonia, Nicholas Théry, President of Crédit Mutuel, Yeong Zee Kin, Assistant Chief Executive, Infocomm Media Development Authority of Singapore, Sasha Rubel, Programme Specialist, Knowledge Societies Division, UNESCO. See my panel speaking points below:
AI BRINGS INCREDIBLE POTENTIAL
- AI is transformative in healthcare, learning, public services, safety, transport, sustainability…
- For example, MIT professor Regina Barzilay, is developing AI to analyse mammograms to detect cancer earlier, after a screen missed her illness.
- By identifying patterns, AI can also contribute to real-time detection of financial markets fraud.
- AI optimisation can drive more efficient energy use and transportation planning & reduce menial tasks.
- AI is growing: AI start-ups attract around 12% of global private equity investments & AI patents grew 10x between 1990 & 2016 (3x for all patents).[i]
AI BRINGS CHALLENGES: BIAS
- AI systems work based on patterns detected in datasets, so they risk reinforcing existing biases in society.
- Gender biases: A study found, for example, that men were almost six times more likely than women to be shown ads for high-paying executive jobs.
- Racial biases: Research has shown that AI systems sold by tech giants have error rates of max1% for lighter-skinned men and 35% for darker-skinned women.[ii]
AI BRINGS CHALLENGES: PRIVACY, DEMOCRACY
- AI can infringe on human rights and privacy: risk of data manipulation & sharing; identification and tracking; speech and facial recognition. Evidence of AI tech being used by governments to spy on civil society and activists.
- AI can infringe on democracy: algorithms and ‘bots’ help share fake news, violent images, harvest & sell data (cambridge analytica) and drive echo chambers, esp on social media.
GLOBAL MULTISTAKEHOLDER RESPONSE NEEDED
- We need to avoid multiplication of approaches: currently national AI strategies, AI ethical codes by businesses, AI standards from technical community.
- OECD has counted 26 sets of AI guidelines developed by stakeholders.
- No single country or category of actors has all the answers to these challenges. We need a global multi-stakeholder response to a global issue.
- We need to speak the same language & develop sound evidence-based approaches to common AI goals.
OECD HAS DELIVERED AI PRINCIPLES
- This year 36 OECD Members and 7 partner countries have adopted the OECD Recommendation on AI, the first intergovernmental set of AI Principles.
- They served as the basis for the G20 AI Principles.
- The OECD AI Principles represent a global reference for responsible stewardship of trustworthy AI.
- They identify shared values-based principles: sustainable growth & well-being; human-centred values and fairness; transparency & accountability.
- They recommend specific policies for international co-operation, to foster alignment around technical standards, metrics and interoperability.
IMPLEMENTING THE PRINCIPLES: OECD AI POLICY OBSERVATORY
- OECD AI Policy Observatory is a multi-stakeholder, multi-disciplinary hub to provide guidance on implementing the AI Principles, policy analysis and resources, and indicators and measurement work.
- It will include a digital library of countries’ and stakeholders’ AI initiatives to help mutual learning: so far 270+ policy initiatives from 40+ countries.
- Given the global nature of AI governance and thanks to its broad reach and inclusive approach, the AI Policy Observatory will be integral to the international debate on AI policy solutions.
G7 AGENDA: OECD ALSO ENGAGING ON GPAI
- OECD called upon by France & Canada through the G7 to support the Global Partnership on AI.
- GPAI will work with OECD to guide the responsible adoption of AI grounded in human rights.
- There is great scope for convergence between AI initiatives globally: we all recognise the need for fairness, transparency, and accountability.
- This consensus is crucial, but we cannot rest there. We need to raise our ambitions for concrete actions and implementation mechanisms, like the OECD AI Policy Observatory that will facilitate AI global governance in the years to come.