At Rise of AI 2018 in Berlin, Irakli Beridze, Senior Strategy and Policy Advisor for Artificial Intelligence and Robotics at the United Nations, describes how the UN is trying to maximize AI's benefits for the Sustainable Development Goals while managing its risks. Speaking just after the UN's Global Summit on AI for Good in Geneva, he traces how fast military and civilian technology changed in fifteen years, from zero military robots in the 2003 Iraq war to hundreds of thousands or millions in use by 2018, and describes the UN's own path from a first AI discussion at the General Assembly in 2015 to a Centre for AI and Robotics opened in The Hague. He surveys national AI strategies, including China's goal of matching the United States by 2020, leading by 2025 and dominating the field by 2030, and names four risk areas: technological unemployment, with no agreed remedy; AI-enabled crime such as deepfakes; lethal autonomous weapons, where experts agree only that a human should make the final kill decision; and existential risk from superintelligence. He closes with examples in health, education and justice, and calls for closer UN, government, academic and industry cooperation.
Key takeaways
- Beridze contrasts zero military robots in the 2003 Iraq war with hundreds of thousands or millions in use by 2018.
- He says China's AI strategy targets matching the US by 2020, leading by 2025 and dominating by 2030.
- He calls technological unemployment probably the biggest near-term AI risk, with no agreed remedy yet.
- UN experts agree only that a human should make the final decision to kill in autonomous weapons systems.
- He cites a Berlin-based startup giving refugees accredited university education through AI-driven virtual assistants.
Detailed notes
The United Nations' growing engagement with AI
Beridze reports directly from the UN's Global Summit on AI for Good in Geneva, where 32 UN agencies met to explore how AI can help solve problems tied to the Sustainable Development Goals. He traces the UN's institutional path on the issue: a first side event on AI at the UN General Assembly in 2015, which he chaired and which still struck most delegations as science fiction; the UN Secretary-General publicly calling AI a potential game-changer for the first time at a later General Assembly session; the first UN Global Summit on AI for Good; and the opening, last year, of a UN Centre for AI and Robotics in The Hague, which connects member states with stakeholders, runs educational programs for media, policymakers and diplomats, and conducts country assessments to match available technology to a country's development problems.
Fifteen years of acceleration
Beridze contrasts the 2003 Iraq war, when zero military robots were deployed, with 2018, when hundreds of thousands or millions of military robots and AI systems are in use, and pairs this with the shift from 2003-era mobile phones and VCRs, discontinued in 2008, to the smartphone era. He expects the coming decades to bring a human-AI coexistence that societies must actively manage to maximize benefits and minimize harm.
National AI strategies
Beridze names the United States, China and Russia as leading nations, noting more than ten billion dollars in US venture funding for AI and the presence of major companies such as Amazon, Facebook, Google, IBM and Microsoft. He says China's AI strategy, adopted the previous summer, sets out to match the United States by 2020, lead by 2025 and dominate the field by 2030. He also cites the EU, Canada, Japan and the UK as active, the United Arab Emirates for appointing a Minister of Artificial Intelligence, and Saudi Arabia for granting citizenship to the robot Sophia.
Four risk areas
Beridze lists technological unemployment as probably the biggest near-term risk, with no agreed solution among proposals such as universal basic income or retraining; he says he recently launched a UN project with more than 55 countries to study how AI-driven automation will affect social stability in developing and transition economies. He flags AI-enabled crime, including deepfakes and impersonation, as a coming challenge for law enforcement, citing planned work with Interpol. On lethal autonomous weapons systems, he describes a UN Group of Governmental Experts that meets twice a year in Geneva without reaching a solution, though member states agree the final decision to kill should be made by a human, without yet having a way to verify or enforce that. Finally, he names existential risk from superintelligence as a risk to keep in mind while designing systems, even if it is not an immediate concern.
AI for the Sustainable Development Goals
Beridze frames the AI for Good summit as the UN's largest effort to align agencies, private-sector and academic resources behind the 17 Sustainable Development Goals, 169 targets and 230 indicators for transforming the world by 2030. He cites AI's potential to analyze health data, noting that the head of the World Health Organization told the Geneva summit that hopes for AI to predict Ebola outbreaks have not yet been realized, as an outbreak has resurfaced. On education, he mentions a Berlin-based startup providing virtual assistants that give refugees access to accredited university education, and on Goal 16, peace, justice and strong institutions, he cites work with Interpol against data manipulation and fake news. He closes by calling for closer cooperation between the UN, governments, academia and the private sector to carry this work forward.
Quotes
AI probably potentially is the most powerful technology we have ever created.
Irakli Beridze
The final decision to kill should still be made by a human being.
Irakli Beridze
Questions answered in this talk
What does China's AI strategy aim to achieve by 2030?
According to Irakli Beridze at Rise of AI 2018, China's AI strategy aims to match the United States in AI by 2020, be ahead by 2025, and dominate the world in AI by 2030.
What is the UN's position on lethal autonomous weapons systems?
According to Irakli Beridze at Rise of AI 2018, a UN Group of Governmental Experts meets twice a year in Geneva on the issue, and member states agree only that the final decision to kill should be made by a human, without yet having a way to verify or enforce that.
What does Irakli Beridze consider the biggest near-term risk from AI?
According to Irakli Beridze at Rise of AI 2018, technological unemployment is probably the biggest near-term risk from AI, and no proposed remedy, such as universal basic income or retraining, is yet adequate to address it.