SPEECH BY DR JANIL PUTHUCHEARY, SENIOR MINISTER OF STATE, MINISTRY OF HEALTH, AT IHIS-MICROSOFT GENERATIVE AI MICRO-CONFERENCE ON FRIDAY, 7 JULY 2023

Ms Ngiam Siew Ying, CEO, IHiS

Ms Lee Hui Li, Country Managing Director, Microsoft

Colleagues and friends

1.            Good afternoon. Thank you for inviting me to join you.

The importance of AI in Healthcare

2.            There have been amazing advancements in the application of AI in medicine, such as the detection of diseases from medical images, and detection of adverse drug reactions from electronic health records. Many of these things today are routine uses of technology, but it is only in a very short period of time, that these have been improved by the use of technologies such as artificial intelligence (AI).

3.            As a result, AI and the technological tools that we deploy have allowed our healthcare professionals to better care for patients and I am going to cite a couple of examples. I am sure we will have many more as we go through the afternoon.

4.             Together with Health Sciences Authority (HSA), IHiS developed an AI system to detect adverse events related to medicines using electronic health records. The amount of data in electronic health records increases in complexity, depth and volume as the kind of care that we can deliver improves over time. Trawling through that data to be able to then identify what are the previous adverse drug reactions is very important rather than relying on the memory of either the healthcare professionals or individual patients. As a result, our regulator, HSA has been able to proactively monitor the safety, in particular recently, of COVID-19 vaccines approved in Singapore, highlighting issues and insights for medicine safety surveillance. This has allowed us that speed of communication and engagement with the public to reassure them that indeed the COVID-19 vaccines were safe in near real time, something that I don’t think we have ever been able to do with any other vaccination programme in healthcare. That was an important outcome for us as a society.

5.            Another example, IHiS and Changi General Hospital have developed an AI Engine that can accurately predict the likelihood of a patient having mild or severe pneumonia based on their chest X-ray image. For the lay public, surely a doctor can just look at the X-ray and decide if it is mild or severe. As a former clinician, this is something of debate because X-rays like other parts of your information space and healthcare, evolve over time. Can you detect the severity of the pneumonia, before it becomes obvious? The use of large volumes of data and particularly the imaging data that is available through X-rays allows us that kind of examination which the human eye and the human mind cannot do in real time. And that ultimately helps healthcare workers identify patients who are at risk of becoming critically ill, allowing interventions to happen earlier and better. Generative AI has changed the conversation.

Generative AI marks a substantial leap forward in AI system capabilities

6.            Recent breakthroughs have changed the way we think about what AI can do and should do. Traditional AI has focused on detecting patterns from images and understanding natural language. Generative AI, on the other hand, produces new content, such as audio, text, and images.

7.            The engagement of these tools with the data sources has gone beyond singular data modalities. Now many of these platforms are processing, not just texting and responding to visual inputs and integrating a variety of data modalities. This marks a substantial improvement in AI capabilities.

8.            In essence, these latest systems can see, hear and appear to “understand”,  but you all know that the gap between what appears to be human understanding and what is in fact, human understanding is still a gap that is quite large. Ultimately, these AI systems interact with the world in increasingly complicated and sophisticated ways.

Safe and responsible use of AI is important

9.            These advanced models are still flawed and limited. They display a bias and they potentially can overstep the guardrails intended to prevent offensive or harmful statements. Some of them actually fabricate false information, which we have euphemistically coined the term “hallucinations”. As someone who had to study psychiatry as part of medical school, I find that a little bit worrying, but it is a term that we have chosen to use and we will say that the AI is hallucinating.

10.       As a result, we have to find ways to prevent real harm that can arise from the flaws in the system. We are committed to make sure that AI is safe and we have to instill confidence and trust among users. I think all the stakeholders appreciate that unless we continually build and maintain that trust, ultimately, any tool no matter how glamorous, how fast moving and how agile, is not going to be able to be adopted unless the users have that base of trust. We have to work on that basis of instilling confidence and trust amongst the users. We are not starting from scratch in advocating for safe and responsible AI. We have had some initiatives that we have been working on. Let me mention three examples.

11.       HSA oversees AI Medical Device (AI-MD) development to ensure its quality, safety, and efficacy. They also require post-market surveillance to ensure continued effectiveness. This is not a simple licensing regime. There are a lot of granular technical work as a regulator that they put in to assess the deployment and operation of these tools in the real world.

12.       The Ministry of Health (MOH) has produced a guide for developing healthcare-related AI, the AI in Healthcare Guidelines, for developers and implementers, identifying what are good practices that they should adopt in rolling out such clinical services to their patients. It is a baseline for how we can develop a common mindset around implementation, policies, regulations and guidelines.

13.       The Infocomm Media Development Authority (IMDA) has introduced AI Verify, which is a AI Governance Testing Framework, so that companies that wish to demonstrate that their AI tool is indeed responsible, could do so in an objective and verifiable manner. IMDA has launched the AI Verify Foundation and it allows anyone to contribute and promote the growth of new and better AI testing tools.

14.       We can do more in promoting a safe and responsible AI approach. By working together, the Government, industry and research communities can develop and implement AI solutions that meet ethical and governance standards outlined in our AI governance frameworks and guidelines.

Partnership between IHiS and Microsoft

15.       This is essentially what today’s conference is about, bringing together the various communities to think through how do we do the implementation, development and deployment work. In addition to enhancing our knowledge about generative AI and its practical use in public healthcare, an event like this serves as an opportunity to spark interest among our partners and encourage collaborative efforts in driving AI innovations.

16.       As part of the conference, we will also witness IHiS and Microsoft signing a memorandum of understanding (MOU) to further explore the opportunities to advance the development and use of generative AI applications.

17.       Together, IHiS and Microsoft can advance health AI innovation meaningfully by combining IHiS' capabilities in developing and managing healthcare IT systems, with Microsoft's expertise in advanced AI technologies and both having a strong commitment to privacy and security.

18.       One collaboration under this MOU is for IHiS and Microsoft to build a Minimum Viable Product for a “Secure GPT for Healthcare Professionals”. I look forward to hearing more details about this later on.

19.       This could serve as baseline for public healthcare institution partners to explore and develop innovative use cases. These can range from information chatbots to clinical notes summarisers to potentially tools that can transform clinician workflows through automation.

Closing

20.       It is through these types of collective efforts, like the one we are commemorating today, that we can address the challenges that lie ahead. We will work, learn, and grow together, to ensure that AI is not just a tool of progress but also a testament to our commitment to safety, ethics, and responsible innovation; and we must all reaffirm our commitment to the safe and responsible development of AI.

21.       Once again, congratulations to IHiS and Microsoft on this milestone partnership.

22.       I wish you all a productive and enlightening conference ahead. Thank you.