AI in general 1, 2018

National Strategy for AI - Discussion Document

India

Actors

National Institution for Transforming India (NITI Aayog)

Tags

Bias, Fairness, Labour market impacts, Privacy,

Resources


In the fiscal year, 2018-19 India is launching a National Program on AI to guide R&D in this domain. The National Institution for Transforming India is pursuing multiple activities around AI, which include working towards a national strategy, running proof-of-concept studies in critical areas like health and agriculture, and collaborating with experts and stakeholders. The two latter activities support the creation of a national strategy, about which a first discussion paper was released in June 2018. 
The document is meant to lay the groundwork for an evolving strategy and puts forward the #AIforAll brand, which seeks to position India as a global leader in the field. The brand suggests an inclusive approach, in line with the country’s development goals. AI is viewed as a ‘once-in-a-generation’ opportunity, the benefits of which should extend beyond financial impact and support the greater good. The report suggests that India may be an ideal ‘playground’ for global enterprises to develop scalable solutions for the rest of the world’s developing and emerging economies. These regions share a commonality of issues that need to be solved, hence, ‘Solved in India’ could act as a seal of approval and become a model for developing AI as a Service for other economies.  The report thus proposes making India an ‘AI Garage for 40% of the world’.
The document focuses on 5 key sectors where AI could bring the greatest societal benefit:
[ul]
  • Healthcare;

  • Agriculture;

  • Education;

  • Smart cities and infrastructure;

  • Smart mobility and transportation.

  • [/ul]
    In addition, cross-cutting barriers are also identified, which could hinder AI deployment at scale, and thus call for an integrated approach:
    [ul]
  • Lack of enabling data ecosystems;

  • Low intensity of AI research, both core and applied;

  • Inadequate availability of AI expertise;

  • High resource cost and low awareness of adopting AI;

  • Unclear privacy, security and ethical regulations;

  • Unattractive IP regime.

  • [/ul]
    Recommendations to address these barriers are put forward in 4 categories:
    [ul]
  • Research;

  • Skilling for the AI age;

  • Accelerating adoption;

  • Ethics, privacy and security.

  • [/ul]
    With regard to research, a two-tiered structure is proposed to address the fact AI research in India is still in its infancy:
    [ul]
  • Establishing Centres of Research Excellence (CORE) dedicated to core research;

  • Establishing International Centers of Transformational AI (ICTAI) to develop and deploy application-based research in collaboration with the private sector.

  • [/ul]
    COREs would focus on basic research and act as ‘technology feeders’ to ICTAIs, dedicated to creating AI-based applications and accelerating adoption in key domains.
    In addition, it is suggested the Government of India should take the lead in bringing together parties to create ‘People’s AI’, an initiative similar in scale and scope to that of CERN. However, unlike CERN, such a centre would not need a physical location and could be distributed across the world, with India acting as the initial funding and coordinating agency. The centre’s mandate should be to focus on cross-cutting, foundational issues that can make AI inclusive, such as General Artificial Intelligence; explainable AI; advanced anonymisation protocols; ethics in AI; and leveraging AI in the service of the world’s biggest problems in health, education, agriculture, etc. 
    With regard to the skills shortage, the document proposes a set of interventions for students and another set aimed at the workforce.
    Measures should be taken incentivise job creation in the new service industry, ideally, roles that are part of the AI solution development value chain but require low levels of expertise, such as data annotation, image classification, speech transcription, etc. These can provide employment at scale. In addition, the quality of informal training institutions that offer to reskill workers into technology-related professions should be standardised. Open online platforms for self-learning - similar to Coursera and edX - should also be created and their offerings standardised through certification and quality measurement. Finally, models for co-funding reskilling by government and companies should be explored. This may take multiple forms, such as income tax deductions, special taxes to be paid if a training budget is not distributed, or grants subsidizing training.
    The interventions addressing students are broad and encompass basic reforms to the whole of the Indian education system, including the introduction of skill-based learning in subjects relevant to AI, increasing the amount of project related work across education levels, and increasing collaboration between industry and academia. To address the lack of qualified faculty, credit-bearing MOOCs and other decentralised teaching mechanisms should be explored. Bridge courses at the postgraduate level should be established for non-computer science students who have other domain expertise. 
    Establishing a task force to monitor changes in employment caused by AI in India is suggested as a solution for creating a longer-term, sustainable framework.
    The document describes how India is lagging in AI adoption, despite its strong presence in the IT industry. Therefore, government involvement seems necessary to promote AI adoption. This extends to private enterprises, public sector undertakings and the government itself. Several recommendations are advanced to address this issue.

    1. A decentralised, blockchain-based multi-stakeholder National AI Marketplace (NAIM) is proposed, that would allow businesses to offer various parts of the AI value chain as a stand-alone service, from data capture to the deployment of solutions. It is hoped that such a marketplace could level the playing field and address information asymmetry while incentivising and simplifying collaboration between the various stakeholders in the AI ecosystem. According to the report, all businesses, government agencies, startups, and research institutions would sign up to the marketplace and engage in their respective activities. Its 3 elements would be a data marketplace, a data annotation marketplace and a deployable model/solutions marketplace. The government doesn’t intend to build such marketplaces itself, rather, to create enabling regulations for private players to do so. This would entail creating standards for personal data use, anonymisation, annotation accuracy and cybersecurity.

    2. The government could assist the creation of large annotated national datasets to spur research and innovation.

    3. Partnerships and collaboration should be supported, especially collaboration between research organisations (AI+X), where AI researchers work with other domain experts, as well as collaborations between research and industry, trade bodies, and venture capital.

    4. Low visibility of AI research and its benefits is a hurdle, which could be overcome by a government managed online AI Database with information about existing projects, networks and results, as well as expert profiles. In addition, efforts at spreading awareness among government agencies and the public sector are also recommended.

    5. Supporting AI startups through dedicated incubation hubs in collaboration with State Governments and private sector stakeholders, and establishing a fund to provide grants to startups.


    With regard to ethics, security and privacy, the document discusses several issues. On fairness and biases, it suggests a reactive approach, identifying biases built into systems and assessing their impacts in order to reduce them, until techniques are developed to create neutral solutions. The strategy acknowledges the importance of the issues around transparency and suggests aiming at explainability, which must be balanced against the risk of attempts at ‘gaming the system’. On privacy, the strategy takes the position that the main worry about companies harvesting large amounts of consumer data and deriving insights is that consumers themselves don’t have access to such insights or derive value from them. Therefore, it suggests that beyond compliance “companies can consider how to create awareness of how they use consumer information and the value they provide in return, which can build trust in their brand and services.” However, the authors of the strategy disagree with concerns that companies with large datasets could build an unfair competitive advantage and they see no negative impact on the consumer.
    In order to deal with privacy issues a number of concrete recommendations are made:
    [ul]
  • Establish a data protection framework with legal backing. The report references the core principles of the proposed Srikishna Committee data protection law: informed consent, technology agnosticism, data controller accountability, data minimisation, holistic application, deterrent penalties and structured enforcement;

  • Establish sectoral regulatory frameworks;

  • Benchmark Indian data protection and privacy laws with European standards like the GDPR;

  • Encourage Indian enterprises to adopt international standards like those of the IEEE;

  • Encourage self-regulation and the use of Data Privacy Impact Assessment Tools;

  • Invest in research to find new mathematical models for preserving privacy;

  • Make citizens aware of their right to privacy, which has been termed a fundamental right by the Supreme Court of India.

  • [/ul]
    The report considers security and accountability to be linked issues and suggests a framework to handle it. The framework would rely on self-regulation of stakeholders to conduct damage impact assessments at every stage of development, and use negligence tests for damages caused by AI. Liability would be limited if appropriate steps to design, test, monitor and improve products had been taken.
    It is suggested that a consortium of Ethics Councils be set up at each Centre of Excellence to define standards of practice.
    Finally, inspired by the UK’s Centre for Data Ethics and Innovation, the document advises the Indian Government to set up a Centre for Studies on Technological Sustainability (CSTS) to address issues relating to ethics, privacy, legal aspects, social sustainability and global competitiveness.
    The document provides no roadmap towards implementation, nor details about the budgetary requirements of the proposed recommendations.

    AI Governance
    Database

    This database is an information resource about global governance activities related to artificial intelligence. It is designed as a tool to help researchers, innovators, and policymakers better understand how AI governance is developing around the world.

    We're collecting more examples as we find them. If you'd like to add yours to this map, please suggest an entry.

    Suggest an entry
    • Filter by:
    • Current filters:
    256 entries are listed below
    • Sort by: