INSITE - Singapore - July 2024
- sukhwinder21
- Aug 1, 2024
- 2 min read
Singapore has now launched the Model AI Governance Framework for Generative AI ("MGF-Gen AI")
Summary: As a regional leader in the field of Artificial Intelligence (AI) Singapore has played an important role in setting the course for the development of AI while managing its associated risks. On May 30, 2024, Singapore’s Infocomm Media Development Authority (IMDA) introduced the Model AI Governance Framework for Generative AI (Generative AI Framework). Developed by IMDA and its non-profit subsidiary, the AI Verify Foundation the framework aims to address concerns about generative AI while promoting innovation.
In Detail: The IMDA have released the updated Model AI Governance Framework for Generative AI. Originally published in 2019 and updated in 2020, the current version, released for public consultation in January 2024, addresses risks like hallucinations and copyright infringement associated with generative AI. It offers a systematic approach to these issues while promoting innovation and involving stakeholders such as policymakers, industry professionals, researchers, and the public.
The Framework highlights nine key aspects:
Responsibility: Assign responsibility based on each stakeholder's control over generative AI development.
Data Protection: Use existing personal data protection laws, Privacy Enhancing Technologies (PETs), and ensure high data quality.
Model Evaluation: Assess AI models during training using techniques like Reinforcement Learning from Human Feedback and Retrieval Augmented Generation, and through benchmarking and red teaming.
Incident Management: Set up processes for timely incident reporting and resolution.
External Audits: Conduct external audits using standardized methods to improve transparency.
Security-by-Design: Implement safeguards such as input filters and digital forensic tools for generative AI.
Digital Watermarking: Use digital watermarking and cryptographic solutions to manage realistic synthetic content.
Safety Techniques: Recognize the limitations of current safety methods and use reinforcement learning from AI feedback and post-training evaluations for better control.
Public Empowerment: Promote democratized access to generative AI to enhance public services, workforce development, and sustainability.
Conclusion: As generative AI evolves, global collaboration on policy is crucial. The GenAI Framework aims to start a global discussion on addressing concerns while promoting innovation. It emphasizes accountability, transparency, fairness, robustness, and security.
The GenAI Framework is a first step toward creating a trusted generative AI ecosystem. Based on its 9 dimensions, the Singapore government will work with key stakeholders to refine these guidelines and resources, balancing safeguards with opportunities for innovation.
Comments