The Fifth Elephant 2024 Annual Conference (12th &13th July)

Abstract:

As AI continues to revolutionize various sectors, it brings both unprecedented opportunities and significant risks. In India, sectors such as Agritech, Fintech, Edtech, public services, and Healthtech are rapidly adopting AI technologies. However, the lack of robust risk mitigation strategies can lead to unintended consequences, including data breaches, algorithmic biases, and systemic vulnerabilities.

Over the past year, we have spoken to practitioners and researchers and collated research on the unique risks in each sector and spicific mitigation strategies. This talk outlines these risks and mitigation strategies as well as best practices for regulatory compliance.

Outline:

The session will be in an open discussion with 4-5 panelists. The structure will broadly cover the following:

• Introduction: Overview of AI adoption in Agritech, Fintech, Edtech, public services, and Healthtech in India.
• Risks in AI Deployment: Identification of key risks associated with AI in each sector.
• Mitigation Strategies: Presentation of effective risk mitigation strategies specific to each sector, with an emphasis on practical implementation.
• Policy and Compliance: Discussion on the role of policymakers and compliance teams in implementing these strategies.
• Best Practices: Sharing best practices and guidelines for developers, researchers, and startup founders to mitigate AI risks.

---
title: AI and Risk Mitigation
markmap:
colorFreezeLevel: 2
maxWidth: 300
initilExpandLevel: 3
---
## AI Risks

- Privacy
- Bias
- Hallucinations
- Unclear processes for oversight and audit
- Data quality
- Lack of explainability
- Data governance
- Lack of standardization
- Lack of regulations and accountability

## Societal implications

- Increase in inequality
- Increase in surveillance
- Over-dependence on AI
- Lack of grievance redressal mechanisms
- Mis- and disinformation
- Reinforcing stereotypes and prejudice
- Increase in the digital divide
- Increased profiling

## Mitigation Strategies
- Adopting a shift-left privacy posture
- Multi-stakeholder participation in policy consultations
- Documentation of processes, checks, and testing
- Use case specificity
- Regular vulnerability checks and auditing
- Human-in-the-loop systems
- Ensure regulatory compliance
- Testing for and mitigating common risks during development stage
- Ensure consensual data collection and data minimisation

Participants:

Developers, researchers, policymakers and analysts, startup founders, risk and compliance teams.

Key takeaways:

The discussion drew from the example of the healthtech sector in India to then discuss some of the risks and challenges in the overall industry.

1. Scope for implementation: The acute shortage of doctors, labs, and testing facilities in India highlights the potential benefits of AI in early identification and critical care.

2. Digitization and data utilization: A significant amount of data and a strong focus on digitization across medical centers have led to substantial improvements in the quality of care in Tier 2 and Tier 3 cities.

3. Risks and challenges:

   • Language barriers: There is a major challenge due to the lack of support for Indic languages in current AI systems.
   • Policy gaps: There is a notable gap in policies regarding the handling of vast amounts of medical data.
   • AI bias: AI systems can exhibit inherent biases. To mitigate this, active learning frameworks should be used, and doctors must verify AI outputs to ensure accurate diagnoses.

4. Regulatory and transparency needs:

   • Grievance redressal: Effective and accessible grievance redressal systems are necessary.
   • Report source transparency: Patients should be informed if a report is AI-generated and have the option to request a doctor’s review of the report.
   • Undefined AI risks: The undefined nature of AI risks complicates the creation of clear regulations and liability standards.

5. Education and monitoring:

   • User education: There is a need to educate users about the potential pitfalls and risks associated with AI.
   • Quantifying risks: Due to the black box nature of Generative AI, it is crucial to quantify risks and accuracy through methods such as codification.
   • Geopolitical considerations: Monitoring the geopolitical impacts on AI is important for developing comprehensive regulations.