• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

Think20 suggestion on topic "AI for Inclusion: A Public G20 AI Platform"

Think20 (T20) is an official Engagement Group of the G20. It serves as an "idea bank" for the G20 by bringing together think tanks and high-level experts to discuss policy issues relevant to the G20. T20 recommendations are synthesized into policy briefs and presented to G20 working groups, ministerial meetings, and leaders’ summit to help the G20 deliver concrete policy measures. 

Every year scientists all over the world suggest their ideas for the "idea bank". In 2024 the "AI for Inclusion: A Public G20 AI Platform" was suggested. International team of authors in alphabetical order includes:

Irina Yarygina. Dr., Prof., Head of Economics and Banking Department, MGIMO University.

Jaya Josie. PhD in Public Finance and Administration from the University of the Western Cape; Adjunct Professor at the Universities of Venda and the Western Cape; Advisor to Zhejiang University’s International Business School’s China Africa Center.

Krish Chetty. PhD Candidate in Computing Sciences at Nelson Mandela University; Senior Research Manager, Equitable Education and Economies Division, Human Sciences Research Council.

Maria Yudina. PhD. in Sociological Sciences, Senior Research Fellow at the Laboratory for Digital Transformation in Public Administration, Institute for Public Administration and Governance of HSE University.

Nirmala Gopal. Professor in Criminology, University of Kwazulu Natal.

Manjeet Kripalani. Executive Director, Gateway House.

Sanjay Anandaram. Ambassador, Governing Board Member, iSpirt.

Wu (David) Wen. PhD in Computer Science from Oxford University; Executive Deputy Director, Fintech Security International Research Center, Zhejiang University’s International Business School.


AI for Inclusion: A Public G20 AI Platform

Artificial Intelligence (AI) is a highly transformative technology offering developmental opportunities for marginalized communities across developing and emerging countries. However, the costs of AI access through private platforms are becoming exclusionary, negating these opportunities. AI development and advancements depend on three inputs: algorithmic ingenuity, computing power and input data. Private AI companies, largely Western, have invested heavily in these three, offering commercial services at a fee. They are near oligopolistic, keeping the benefits out of reach of the ordinary citizen. It is critical to make the advantages of AI available to all. To create this equity, the G20 nations have an opportunity to make similar investments and launch a public offering benefiting all citizens. 

An enabling platform providing transformative AI services can be the public good that the G20 espouses. Such offerings can complement the private sector’s offering and must coalesce the three inputs for AI: algorithms, computing power and data. 

For algorithms, the G20 can organize around the Open Source Software Development Community. An example is India’s Code for GovTech (C4GT), which demonstrates the potential of mobilizing a large talent pool to run complex models and projects, ensuring continuous innovation and sustainability. These models must be designed keeping in view inclusive principles and incorporate the multiple languages and cultural dimensions of traditional societies especially when developing solutions for critical sectors like agriculture, education, health and justice. 

To add computing strength, the G20, through public-private partnerships, can invest in and launch AI-focused data and computing centers in developing countries outfitted with the necessary hardware to support mass access to a public AI platform. This infrastructure can be creatively powered by green energy, in keeping with the G20’s environment and climate goals.

For the model’s input data, the G20 must adopt data-sharing principles to develop custom models informed by national datasets; this will ensure the models overcome inherent biases. The model must respect national data privacy laws, promoting the training of AI across multiple data sets without necessarily moving the data across borders.  To this end, the model applies a creative decentralized "hub-and-spoke” data management structure, allowing nations to control their data sovereignty.

We propose an AI Stack, an evolved concept based on the India Stack, which leverages India’s well-coordinated Open-Source Software Development community, offering a series of public digital services to both marginalized communities and the commercial sector, levelling the playing field for all. The AI Stack will be a comprehensive set of AI services specifically designed to transform the informal sector and give small business owners access to advanced business functions to enhance their competitiveness. 

AI-driven tools can automate and streamline HR processes, personalize marketing strategies, and provide predictive analytics for more informed decision-making across governance, IT, legal, and financial operations. AI can optimize supply chains, predict market trends, and enhance customer engagement for procurement, operations, and sales. In quality assurance and compliance, AI can continuously monitor and ensure adherence to standards and regulations. These uses are invaluable for the informal sector, which has had difficulty in digital adaptation. Using these advanced tools will facilitate a smoother transition into the formal sector, a G20 goal for sustainable growth and competitiveness.

India Stack’s success lies in its ability to leverage the trio of community-driven initiatives, government support, and market innovations to create digital public goods at population scale, providing open, scalable, and inclusive services as exemplified by platforms like Aadhaar, CoWIN, and DIKSHA. The community contributes through open-source software development, providing innovative solutions and enhancements. The market innovates on these platforms, creating customized AI services for various sectors like HR, marketing, and finance.

By adopting this model, the AI Stack can benefit from a rich ecosystem of developers, entrepreneurs, and users, all contributing to a common goal. The open-source nature ensures the tools are accessible, adaptable, and cost-effective, making them ideal for the informal sector. This approach democratizes access to advanced technology and aligns with the T20’s interest in inclusive digital transformation. 

The G20, as a collective of the world’s major economies, possesses the unique capacity to coordinate centrally data collection from the informal sector, thereby ensuring that open-source software developers are creating solutions that directly address the sector’s challenges. Using a data repository built upon a "hub-and-spoke" model, the G20 can gather, anonymize, and analyze vast amounts of data from various informal sectors across member countries. This allows decentralized control over national data. This data can provide invaluable insights into the specific needs, behaviors, and challenges of informal businesses, offering a granular understanding of country-specific and shared experiences to drive the development of targeted solutions. 

To facilitate this, the G20 can initiate a standardized framework for data collection, ensuring consistency and comparability of data. Collaborating with local governments, NGOs, and industry bodies can enhance data collection efforts, and the data can then be made available to the open-source community under strict ethical guidelines and privacy standards. 

For the AI stack to be effectively accessible at scale by the collective informal sector of G20 countries, the G20 must mobilize substantial digital infrastructure resources. It entails establishing a decentralized "hub-and-spoke" model where the central hub is equipped with high-speed internet access, fundamental for seamless connectivity and real-time data processing. Cloud computing resources at scale will provide the computational power and storage capacity required to run sophisticated AI algorithms and handle large datasets characteristic of diverse informal sectors. Cybersecurity protections are crucial to safeguard sensitive business and personal data against increasing cyber threats, ensuring trust and reliability in the AI stack. Additionally, robust data storage and management capabilities are required to efficiently handle the influx of data, ensuring its availability, integrity, and confidentiality. Stringent data protection laws already exist, like GDPR and the European AI Act. India has the Data Protection and Privacy Act of 2023 (DPDP Act) and a techno-legal data-sharing approach called Data Empowerment and Protection Architecture (DEPA). This is being extended to training AI models and can detect fraud across multiple data banks. It promotes transparency in digital contracts, "confidential clean rooms" with hardware-protected secure environments, and an effort to create an open and fair ecosystem.  

Beyond the technical infrastructure, the G20’s AI efforts are sure to attract a dedicated group of professionals from across the G20 countries, capable of coordinating the efforts of the open-source software development community. This will require a pooling of resources, knowledge, and expertise from all member countries, alongside strategic partnerships with technology firms, academic institutions, and non-governmental organizations. 

Creating this will require the energies of the three G20 troika countries, all developing countries with technical and financial talent and populations with need. If it is successful, a pilot between India, Brazil, and South Africa can be attempted and rolled out to takers in the G20. If it succeeds, it will be a truly collaborative G20.