Leveraging AI Potential By Ensuring Adoption of Responsible AI

Realizing the need of the hour to leverage the vast potential of Artificial Intelligence (AI) for social and economic benefit while mitigating the risks, the Working Group on Responsible AI (RAI) of the Global Partnership on AI (GPAI) presented the findings of several initiatives at the ongoing GPAI Summit.

In a session moderated by Sophie Fallaha, Executive Director at the International Centre of Expertise in Montreal for the Advancement of Artificial Intelligence (CEIMIA), the members presented the update and key learnings from the six key initiatives.

The Working Group on RAI is currently pursuing six initiatives: Responsible AI Strategy for the Environment (RAISE), Scaling Responsible AI Solutions, Towards Real Diversity and Gender Equality in AI, Responsible AI for Social Media Governance, Sandbox for Responsible AI – A Procurement Scenario and AI-Powered Immediate Response to Pandemics.

Sandbox for Responsible AI

Commenting on the learning from the Sandbox for Responsible AI initiative, Aditya Mohan, GPAI Project co-lead sandbox for responsible AI, said, “In 2023, we reviewed a whole spectrum of existing AI procurement resources. In the coming year, we plan to design and implement a procurement sandbox environment to test these best practices in a public procurement use case developed in collaboration with a government based on the needs. This would be followed by publishing guidelines and best practices for AI procurement sandboxes in 2025.” 

Scaling Responsible AI Solutions

A key component of leveraging the benefit of AI is to scale the solutions in a responsible manner. “We are working with the aim to provide guidelines to AI teams and policymakers on how to address the challenges that hinder the scalability of responsible AI solutions. We are also learning from teams across the world who have deployed AI solutions and want to scale in a responsible manner,” said Amir Banifatemi, GPAI Project co-lead, Scaling Responsible AI Solutions and Co-Founder and Director of AI Commons.

Responsible AI for Social Media Governance

Another area where AI is playing a disruptive role is social media. Around 4.89 billion people use social media platforms across the world now and thus, it is not surprising that social media platforms are a key vehicle for AI influence on people today.

Speaking about the role of AI in social media governance, GPAI Co-lead on Social Media Governance, Alistair Knott, spoke about three key types of AI systems, including recommender systems, harmful content classifiers and the new generative AI tools, in the social media context.

While elaborating on the harmful effects of AI-generated content, Knott mentioned, “What is needed is a reliable way of detecting AI-generated text. The current tools to detect AI-generated content don’t really work. We believe the way to make them work is to place an obligation on the company. A company that produces an AI-generator must demonstrate and supply, for free, a reliable detection tool for the content it generates, as a condition for its public release.”

Responsible AI strategy for the environment (RAISE) and biodiversity preservation

A pertinent issue of our times is an urgent need to prevent climate change and address the growing environmental issues. AI

“AI has tremendous potential to help better deliver on climate action and biodiversity preservation,” said Nicolas Miailhe - GPAI co-lead RAISE - Founder and President of The Future Society. Speaking on the future vision of RAISE, Miailhe, who presented along with Lee Tiedrich, co-lead RAISE and from Duke University, said, “Do we need to model planetary systems at a high level of granularity in the years to come to deliver on the ecological transition? Yes, there are models like that. Do we need to do that in a manner that renders these models more interpretable…they are like big black boxes right now. For us to break into these big black boxes requires investment, among other things.”

Towards Real Diversity and Gender Equality in AI

Key activities of the sub-group on promoting diversity and gender equality in AI included conducting literature review, regional consultations, building community perspectives, promising practices and resources and environmental scan.

Elaborating on the outlook for 2024, Benjamin Prud’homme, Vice-President, Policy, Society and Global Affairs, said, “2024 will definitely be a bigger year for us as we would compile the results of the 2023 activities and carry additional consultations as required. We would also be developing and deploying a diversity and gender equality toolkit, which is a curated and annotated toolkit that we have identified. It will also include the advantages and limitations of each tool. Lastly, we would also be working on the GPAI report, which will include recommendations for the Governments related to gender equality.”

AI-Powered Immediate Response to Pandemics

The sub-group on building pandemic resilience was formed with the intention to support AI initiatives to help fight against the COVID-19 pandemic. 

“The models that we have created now have helped us build resilience so we can use the right model at the right time. In New Zealand, they didn’t work so well because we didn’t have so many cases but they worked really well in other places like the UK. They can also be used to acquire insights to take better decisions,” said Michael O’Sullivan, GPAI Project co-lead of the Pandemic Resilience working group. He is also an Associate Professor in the Department of Engineering Science at the University of Auckland.