Ensuring Development and Maintenance of Ethical Artificial Intelligence Standards
Ministry of Digital Development and InformationSpeakers
Summary
This question concerns the development and maintenance of ethical artificial intelligence (AI) standards following the emergence of generative models like ChatGPT and GPT-4. Dr Tan Wu Meng inquired about strategies to ensure AI aligns with human values and maintains ethical rigor during development, training, and deployment. Senior Minister of State Dr Janil Puthucheary highlighted the Model AI Governance Framework, the AI Verify toolkit, and upcoming Advisory Guidelines under the Personal Data Protection Act. He explained that public officers remain accountable for vetting AI output and noted that "proof of human" markers for AI interactions are worth studying for specific applications. Senior Minister of State Dr Janil Puthucheary also emphasized using the Advisory Council on Ethical Use of AI and Data to promote accountability and best practices across the private sector.
Transcript
16 Dr Tan Wu Meng asked the Minister for Communications and Information considering the advances in artificial intelligence (AI) with platforms such as ChatGPT and the GPT-4 architecture, what attention is being given to promoting and ensuring (i) the development of ethical AI which responds in a manner consistent with human values and (ii) maintenance of ethical standards in the development, dataset training and deployment of AI.
The Senior Minister of State for Communications and Information (Dr Janil Puthucheary) (for the Minister for Communications and Information): Madam, Singapore supports the responsible development and deployment of artificial intelligence (AI), so that its benefits may be enjoyed in a trusted and safe manner. Our approach was explained at the 19 April 2023 Sitting.
Among other measures, we introduced the Model AI Governance Framework and AI Verify, a self-testing toolkit to demonstrate responsible deployment of AI. Later this year, we plan to issue the Advisory Guidelines on the Use of Personal Data in AI Systems under the Personal Data Protection Act. We regularly engage industry and international counterparts, such as through our Advisory Council on Ethical Use of AI and Data and the Global Partnership on Artificial Intelligence, to keep abreast of developments.
Where necessary and useful, we will update our measures to take into account the impact of developments like Chat Generative Pre-trained Transformer (ChatGPT) and GPT-4. For example, the Public Service has introduced guidelines for public officers using similar technologies to draft documents. These guidelines make clear that public officers are accountable for their work and are responsible for fact-checking and vetting AI-generated content. The guidelines also aim to safeguard data security by reminding officers not to input sensitive information into these applications.
Mdm Deputy Speaker: Dr Tan Wu Meng.
Dr Tan Wu Meng (Jurong): Deputy Speaker, I thank the Senior Minister of State for his answer. I have got two supplementary questions. The Senior Minister of State, in his example of the Public Service gaining experience with the use of ChatGPT and GPT-4, essentially alluded to the idea of a responsible person when AI is being deployed for a purpose. Can I first ask the Senior Minister of State whether this approach will be encouraged in other settings, including in the private sector? So that, for example, if a firm uses AI to make decisions on human resources, there can be appropriate accountability if the AI results in discriminatory behaviour against job applicants.
Secondly, can I also ask the Senior Minister of State if agencies might wish to study the role of a proof of human marker for certain AI applications and platforms? This is so that Singaporeans and members of the public interacting with an online portal can know whether the chat is being generated by an AI or by a fellow human being.
Dr Janil Puthucheary: Madam, I thank the Member for the questions. The role of the private sector is important. We have significant representation from the private sector in our Advisory Council on the Ethical Use of AI. We hope through that platform, these best practices as well as the moves that the public sector is making can also influence the standards and approaches that occur in the private sector; and vice versa, as we learn best practices from the private sector for our role and mission in the public sector.
On his second question about the role of a proof of human marker, I think it is something worth studying. But I would caveat that it is not necessarily something that you may want to introduce as a general rule across all AI mechanisms. I think the Member would appreciate there are a fair number of tasks which can be very safely and reasonably automated without the need for demonstrating a human in the loop. In our day-to-day work and our interactions with our many devices, we know that this is the type of technology that is being deployed safely and to good effect.
12.30 pm
Mdm Deputy Speaker: We are out of Question Time. Order. End of Question Time. Introduction of Government Bills.