Conversational AI in low income & resource settings | IGF 2023

Dino Cataldo Dell’Accio

In this analysis, several key points and arguments about AI applications in healthcare, the potential of AI and chatbots in low-resource settings, the concept of trust in AI and digital technologies, and the need to establish frameworks for evaluating the reliability and trustworthiness of AI solutions are discussed.

Firstly, the importance of user identification in AI applications in healthcare is emphasised. The use of facial recognition for digital identity is highlighted as an effective solution implemented for the United Nations Pension Fund. This demonstrates how advanced technologies like AI can be utilised to enhance security and streamline processes within healthcare systems.

Additionally, the potential of AI and chatbots in low-resource settings is acknowledged. The analysis suggests that these technologies have the ability to address resource limitations and reduce inequalities in healthcare access. To support this argument, a blockchain solution designed and implemented for the United Nations Pension Fund is mentioned. The use of blockchain technology can provide secure and transparent data management, enabling efficient delivery of healthcare services in low-resource settings.

The concept of trust is recognised as crucial in AI and digital technologies. It is argued that the public should have confidence in the solutions and entities that offer these technologies. The analysis highlights the importance of not burdening individuals with technological details, but rather fostering trust in the overall solution. Trust is seen as a vital factor in promoting widespread adoption and acceptance of AI and digital technologies.

Furthermore, the need to establish frameworks for evaluating the reliability and trustworthiness of AI solutions is emphasised. The analysis suggests that not all solutions have the same level of reliability, and there is a need to develop criteria for comparing and contrasting different AI solutions. This would enable the identification of trustworthy and reliable solutions that can be implemented effectively. The speaker believes that such frameworks will promote accountability and transparency in the AI industry.

In conclusion, this analysis brings attention to various critical aspects of AI applications in healthcare, the potential of AI and chatbots in low-resource settings, the concept of trust in AI and digital technologies, and the need for frameworks to evaluate the reliability and trustworthiness of AI solutions. It underscores the importance of user identification, the potential of advanced technologies in addressing resource limitations, and the value of trust in fostering widespread adoption. Furthermore, it highlights the necessity of establishing criteria for evaluating and selecting reliable AI solutions, promoting accountability and transparency in the industry.

Olabisi Ogunbase

Digital patient engagement is crucial for maintaining relationships with patients even after they leave the hospital. Platforms like WhatsApp play a vital role in this aspect. WhatsApp is a powerful digital tool that enables ongoing interaction between healthcare providers and patients. It allows doctors, nurses, dieticians, and social workers to provide guidance and answer patient questions. This continuous engagement helps prevent relapses and educates patients about their health conditions. WhatsApp also serves as a platform for passing on education and notices, and as a support system for patients to share ideas and support each other. However, there are some limitations with the WhatsApp platform, such as delays in response and lack of personalization. Implementing AI in healthcare communication, specifically conversational AI, could address these issues and provide real-time, appropriate responses. Collaboration and knowledge-sharing are essential for driving innovation in healthcare, particularly as technology continues to advance. By working together, we can improve digital patient engagement and achieve better healthcare outcomes.

Rajendra Pratap Gupta

Conversational AI is emerging as a promising solution to improve accessible healthcare in low-income and low-resource settings. A study showed that Conversational AI scored 81% in the MRCGP, surpassing human physicians who scored 72%. This highlights the potential of AI to enhance healthcare delivery and bridge gaps caused by the lack of qualified doctors and inadequate healthcare infrastructure. AI in healthcare is aligned with SDG 3 (Good Health and Well-being) and SDG 9 (Industry, Innovation, and Infrastructure).

However, there are concerns about awareness and implementation of Conversational AI in low-resource settings. Some digital health professionals are unfamiliar with its concept and potential applications. This lack of awareness might hinder successful implementation.

Rajendra Pratap Gupta supports using voice-based data through Conversational AI to increase the accuracy and volume of health data, leading to improved healthcare outcomes. Collaboration and a user-centric approach are crucial in AI implementation. Involvement of different sectors, including the private sector, is vital for sustainable business models. The WHO, ITU, and WIPO play significant roles in facilitating AI implementation.

Addressing the digital divide is important, as 2.6 billion people globally lack reliable internet access, hindering effective AI implementation. Efforts should be made to increase internet access in underserved areas.

Education in AI and robotics is necessary, with initiatives in place to develop courses for students and train frontline health workers. This will create a skilled workforce to utilize AI technologies effectively.

The debate on regulation in AI continues, with some advocating for guidelines over over-regulation to maintain flexibility and ethical standards while promoting innovation.

In conclusion, Conversational AI shows great potential in improving accessible healthcare in low-income and low-resource settings. It requires awareness, collaboration, and efforts to address the digital divide and provide education in AI and robotics. Finding the right balance between regulation and innovation is crucial. By addressing these issues, AI can play a significant role in advancing healthcare and achieving the Sustainable Development Goals.

Sameer Pujari

In this analysis, the speakers focus on the transformative potential of technology, specifically conversational artificial intelligence (AI), in addressing existing gaps in healthcare services. They assert that these gaps, particularly in low middle-income settings, can be effectively tackled through the implementation of technology. The argument put forward is that technology, especially conversational AI, serves as an enabler in bridging the healthcare divide.

One important observation made by the speakers is the need for a people-focused, collaborative, equitable, and sustainable approach when integrating technology in healthcare. They emphasize the importance of considering the specific needs of individuals and communities, as well as fostering collaboration between various stakeholders. In addition, they stress the importance of ensuring that the benefits of technology are accessible to all, regardless of socioeconomic status.

The World Health Organization (WHO) plays a crucial role in this conversation by providing guidance and support for the effective implementation of AI in healthcare. The speakers highlight WHO’s efforts in maximizing the value of AI in healthcare through initiatives such as the global collaboration with the International Telecommunication Union (ITU) and the World Intellectual Property Organization. These efforts aim to harness the potential of AI to improve global health outcomes.

Ethics and regulations emerge as important considerations in the implementation of AI in healthcare. The speakers stress the need for ethical approaches to AI development and deployment, ensuring that the technology is used in a responsible and beneficial manner. They also highlight the importance of regulations to provide guardrails and prevent potential misuse of AI. However, it is asserted that regulations should not stifle innovation but instead strike a balance between regulation and technological advancement.

Education and training play a significant role in achieving responsible AI implementation. The WHO offers courses on ethics and governance of AI to promote understanding and ethical approaches among developers, policymakers, and implementers. These courses aim to equip individuals with the necessary knowledge and skills to navigate the complex ethical considerations surrounding AI implementation.

In conclusion, the analysis underscores the potential of conversational AI in addressing healthcare gaps and improving global health outcomes. A people-focused, collaborative, equitable, and sustainable approach is deemed essential in effectively implementing technology in healthcare. The WHO’s guidance and support, along with the development of educational courses, ensure that AI is deployed ethically and responsibly. It is evident that harnessing the potential of AI requires a well-balanced approach that brings together technology, ethics, regulations, and education for the betterment of healthcare systems worldwide.

Mevish Vaishnav

Conversational AI has the potential to revolutionize the healthcare industry by analysing health conversations and generating valuable insights and decisions. This presents an incredible opportunity to gather and analyze health data from billions of people and clinicians, leading to more effective healthcare outcomes. Supporters argue that Conversational AI can be the starting point for generating health AI. By leveraging the power of Conversational AI, healthcare professionals can better understand patient needs and tailor treatment plans accordingly.

Conversational AI also addresses the lack of access to basic health information, particularly in rural areas. Many people living in remote or underserved locations struggle to access crucial information about their health. Conversational AI can bridge this gap by providing easy-to-understand and readily accessible information. Advocates argue that generative AI could eliminate the need for doctors to address basic health problems.

The potential of implementing Conversational AI and generative health AI is widely recognised. However, no supporting facts are provided to elaborate on this stance.

Conversational AI is also seen as a powerful tool in patient engagement and health-related education. The effort required in typing and texting often hinders effective communication between healthcare providers and patients. However, Conversational AI streamlines this process by allowing patients to converse naturally, making them feel heard and fostering a better doctor-patient relationship.

Advocates propose the creation of a global generative health AI group under the stewardship of Dr. Gupta. This group would bring together stakeholders, regulators, policymakers, doctors, hospitals, and frontline health workers to set a direction for all involved. This initiative is supported by the belief that the United Nations, as the largest multi-stakeholder and multilateral body, is in a prime position to facilitate this collaboration. This would promote partnerships and support SDG3 (Good Health and Well-being) and SDG17 (Partnerships for the Goals).

The Academy of Digital Health Sciences is working on a report about generative health intelligence. This report aims to explore the role of generative health intelligence in shaping the future of healthcare. While further details about the report’s content or expected release date are not provided, it is expected to contribute to advancements in healthcare intelligence.

Training and deployment of generative AI in healthcare are emphasized as crucial. Understanding how generative AI works and developing the necessary skills are essential for effectively utilizing this technology. The positive sentiment towards this necessity stems from recognizing the potential benefits of generative AI in improving healthcare outcomes. However, no specific evidence is provided to further support this argument.

In conclusion, Conversational AI has the potential to transform healthcare by analyzing health conversations, delivering information in remote areas, enhancing patient engagement, and facilitating health-related education. The establishment of a global generative health AI group, the training and deployment of generative AI, and the ongoing work by the Academy of Digital Health Sciences highlight the need to fully harness the potential of this technology. Further supporting evidence and details would strengthen the arguments presented.

Shawnna Hoffman

During the discussion, the potential of conversational AI to bridge the healthcare gap was highlighted as a significant advantage. The ability of AI to provide 24/7 assistance and access to healthcare globally, through mobile phones, was emphasized. This can greatly benefit individuals in remote areas or those who may have limited access to healthcare services. The convenience and availability of AI-based healthcare assistance can help address health disparities and provide support to individuals in need.

The combination of AI with blockchain technology was also discussed as an efficient solution during crisis situations. It was mentioned that during the COVID-19 pandemic, an AI chatbot combined with blockchain technology helped locate over 10 billion pieces of personal protective equipment (PPE) within the first 24 hours. This demonstrates the potential of AI and blockchain to rapidly respond to critical needs and find effective solutions in times of crisis.

The importance of fact-checking AI and ensuring its accuracy was emphasized. Even though AI is probabilistic and not always correct, it is crucial to verify the information provided by AI systems. One of the speakers, the president of Guardrail Technologies, highlighted the need to put guardrails around AI and fact-check generative AI to ensure its reliability and accuracy. This point stresses the importance of being cautious and critical when relying on AI-generated information.

The discussion also raised awareness about the issue of internet access and connectivity for AI solutions to be effective. It was mentioned that 2.6 billion people globally lack internet access, which significantly hinders the overall success and reach of AI solutions like chatbots. Ensuring internet access for all individuals, especially those who currently lack it, is necessary to fully harness the benefits of AI and provide equitable access to its solutions.

A holistic approach that considers individual needs, even in remote locations, was emphasized. The experience from an IBM Watson project was shared, where access points were set up in various villages, allowing people to reach these points in half a day and gain access to medical information. This approach recognizes the importance of tailoring AI solutions to meet the specific needs of individuals regardless of their location or resources.

Lastly, the speakers acknowledged the complexity of implementing AI solutions on a wide scale. It was acknowledged that the challenge extends beyond just conversational AI and that the complexity of the problem makes it difficult to implement AI solutions effectively. This realistic perspective highlights the need for careful planning, research, and collaboration to overcome these implementation challenges.

In conclusion, the potential benefits of conversational AI in bridging the healthcare gap, providing 24/7 assistance, and access to healthcare globally through mobile phones were discussed. The combination of AI with blockchain technology was seen as an efficient solution during crisis situations. The importance of fact-checking AI and ensuring its accuracy, considering internet access and connectivity, adopting a holistic approach, and addressing the challenges of implementing AI solutions were all key points discussed during the session. Overall, the speakers expressed optimism about the potential of AI while also acknowledging the complexities and challenges that need to be addressed for its successful integration.

Sabin Dima

Artificial intelligence (AI) is widely recognised as a powerful tool that can replace certain skills, while still acknowledging the importance of human involvement. It is acknowledged that AI can outperform humans in certain tasks, offering greater efficiency and accuracy. Notably, humans.ai, led by the CEO and Founder, has achieved significant milestones in AI development, including creating the first AI counselor for a government and an AI capable of real-time conversations with 19 million Romanians. These accomplishments demonstrate the transformative potential of AI across various domains.

Data traceability and ethics are emphasised as critical considerations in AI development. The CEO’s firm has developed the first blockchain of artificial intelligence to ensure transparency and accountability in AI systems. Additionally, they have contributed to research papers on the ethical implications of AI, emphasising the need to address these concerns.

In the context of healthcare, the CEO argues for a bidirectional approach to AI, aiming to understand people’s problems and provide effective solutions. Emphasising human-like interaction, the CEO advocates for grasping individuals’ problems and urgency. They envision an open innovation platform that fosters collaboration and comprehensive problem-solving.

While technology itself is not the issue, optimising its usage is crucial. The CEO suggests that resources for experimenting with AI projects are readily available to everyone. The focus should be on tackling real-world challenges and driving innovation across sectors.

Furthermore, the CEO asserts that trust can be bolstered in healthcare through the implementation of AI solutions. For instance, the CEO references a project where they cloned a doctor’s voice to send audio messages to patients, enhancing patient care and building trust.

To better understand and regulate AI, the CEO proposes real-world experimentation. By implementing AI solutions in specific regions, regulators can gain insights and make informed decisions on regulations and policies.

The urgency for action and application of AI is evident throughout the discussion. The CEO highlights the readiness of technology and the availability of skilled professionals passionate about AI. Encouraging seizing the opportunities presented by AI rather than merely contemplating its potential is emphasised.

In the conversational AI domain, the CEO suggests making the technology more accessible to underserved populations in low-income areas. By developing efficient models that can run on mobile phones, conversational AI can bridge gaps in healthcare access.

Finally, AI is portrayed as a beneficial tool for employment, increasing productivity and reducing human error. The CEO suggests that AI can supervise performance and mitigate errors, potentially enabling employees to work fewer days while achieving greater results.

In conclusion, AI is a powerful tool capable of replacing certain skills but not humans. The CEO and their firm exemplify the transformative potential of AI across various domains. Ethical considerations, data traceability, bidirectional approaches in healthcare, effective technology utilization, trust-building, real-world experimentation, accessibility, and increased productivity are crucial aspects guiding the application and development of AI. The overall sentiment strongly favours embracing AI to drive positive change in multiple sectors.

Ashish Atreja

Generative AI and AI technologies have the potential to revolutionise the provision of medical care by overcoming the limitations of time and location, extending healthcare access to a larger number of people, irrespective of their physical location. The use of generative probabilistic models in combination with rule-based care plays a crucial role in bridging the gap between scientific treatments and patients’ understanding.

Addressing healthcare inequity requires collaboration and the appropriate use of technology. Inequities exist not only among patients but also among countries, states, and healthcare organisations. Through collaborative efforts and leveraging technology, healthcare can be democratised, ensuring equal access to quality care for everyone.

AI technologies can bridge the digital divide in healthcare. Existing care solutions have the potential to become global solutions if properly validated. Humans play a vital role as transformation agents in bridging this gap, working collectively across silos to ensure inclusivity in healthcare.

Prominent figure Ashish Atreja advocates for a global thought leadership group on generative AI in healthcare. He believes in the power of collective work and engaging with global partners to drive advancements in healthcare systems. Collaborating and sharing knowledge can contribute to the development and implementation of generative AI solutions worldwide.

Conversational AI has the potential to dispel healthcare fallacies by providing accurate and reliable information. However, it is crucial that the technology behind conversational AI is based on validated and trustworthy sources. The FDA has a tiered system for validating health-related technologies based on their potential risk, ensuring their reliability and safety.

To ensure the accuracy and effectiveness of conversational AI in healthcare, an automated or semi-automated governance framework is needed. Currently, there is no specific framework to regulate the validation of conversational AI in healthcare. Establishing such a framework would help maintain the accuracy and credibility of conversational AI, benefiting patients and healthcare providers.

In conclusion, generative AI and AI technologies have the potential to revolutionise healthcare provision, extending care to more people while overcoming limitations of time and location. Collaboration, inclusivity, and validation of technologies are crucial in addressing healthcare inequity and bridging the digital divide. Through collective work, the creation of a global thought leadership group, and the implementation of an effective governance framework, the potential of AI in healthcare can be fully realised, improving outcomes for patients worldwide.

Rajendra Pratap Gupta:
Hi, greetings from Kyoto, and good morning, good evening, and good afternoon, and for some late night. As I start this very important panel discussion on conversational AI in low resource settings and low income settings, let me first give you a perspective on this and how we build up this session. So while we were conceptualizing this very important topic of conversational AI, I did reach out to a lot of my friends who have long time been in digital health, and I must put this through this forum that a few of them weren’t aware of this topic, and which was a big surprise for me. So I think it makes this session all the more important and relevant, because conversational AI is basic digital health. I mean, this is something that we need for the fact that AI is all pervasive, is getting into every aspect of health care delivery, and more than that, what I call is the 80-80-80 rule. 80% of the people don’t have access to health care or qualified doctors. 80% of the areas that we have do not have anything that they can call as health care, and 80% of the problems people have are treatable by probably OTC medications or non-specialist doctors, and that’s where I think our role comes into very importantly. And if you have to talk about affordable and accessible health care, conversational AI is important. While I was serving the union health minister as advisor, I think my boss was very clear, let’s not force doctors to go to rural areas, because they have studied in urban cities for better life, better conditions, and rural areas don’t provide that infrastructure. So even if they go there, what they will do? I mean, coming from that reality, from a country which is of a large population of 1.4 billion, and knowing the effect of what most of these LMICs pass through, I must also relate one experience I had with one of the country heads of IGF who came to our booth and just asked me a question. We’ve been hearing a lot about generative AI. Will it solve our health care problems? And my immediate, instant response was generative AI is based on data. We do not have that. What will it analyze? So if you’re having a very high expectation of saying that generative AI will immediately solve problems, I’m sorry to say that there has to be a baseline of data, there has to be a baseline of clean data for generative AI to work on. So while there is hope and hype, there is a long journey ahead for all of us. With that having said, I also must give you a very interesting example of conversational AI, which is actually chatbots, AI-based chatbots. So in my book, I do mention about this example that there is a very, what do you call, highly respected exam that doctors aspire to pass through. It’s called MRCGP in the UK, member of the Royal College of Physicians. The conversational AI chatbots, they scored 81% compared to human physicians who scored 72%. So I think the evidence is around that there is a future of conversational AI. In fact, there is a present if we deploy it very well, what conversational AI can do. But what we need to create is awareness because so-called LinkedIn leaders of digital health didn’t actually know about it when I reached out to them. They’re all friends. But today, those whom you have on the screen are the actual leaders who understand. Those who are sitting next to me, Dinu and Shauna. So what we are going to do today is to ask people their experience, their expertise and their expectations from conversational AI. And with me, I have Dino Cataldo Dell’Accio, who serves as the Chief Information Officer of the UN Joint Staff Pension Fund and leads the UN Digital Transformation Group. Besides, he has many accolades, but I will just point the one that he got was the UN Secretary General’s Award for his work in applying blockchain technology for digital certification of entitlement process of the UNIJSPF beneficiaries and retirees. Mr. Sameer Pujari, who is among many hats he wears, leads the AI at WHO and is the Chair of AI for Health at WHO ITU focus group. But besides that, he has done a number of things, including Be Healthy, Be Mobile, which is the first mobile app from WHO for chronic diseases. We have Sabin Dima. I think we all in the world of AI and blockchain, I would personally have the highest hope from him given the fact that he’s the first person in the world to merge AI and blockchain together, the founder and CEO of humans.ai. And he’s an entrepreneur who started his first social media at the age of 16. And if you hear him, I bet you that it will change your perspective on what AI and blockchain can do. We have Ashish Atreja. He’s a doctor. He’s a gastroenterologist. He’s currently the CIO and the Chief Digital Health Officer of UC Davis Health. And he’s in pioneering work for digital health. And at least the reason I got him here, he had done phenomenal work during COVID with chatbots, the conversational AI. We have Mavish Vaishnav, the Group CEO of Digital Health Associates, who sits on various government committees for digital health. I’ve been a part of the UN initiative of the Innovation Working Group Asia, where she drafted the roadmap for telemedicine way back in 2013. We have Dr. Olubisi Ongobase, who is a pediatric doctor and quality improvement team lead and mentor. And I’ve seen her work at the World Health Organization. She’s phenomenal and fantastic work that she has done. So what I’m going to do is pass this to my expert panel for their opening remarks on what I have there to say about the conversational AI for low-income and low-resource settings. So I’m going to start with Mr. Dino Cataldo. Dino, over to you for your views on this topic.

Dino Cataldo Dell’Accio:
Thank you very much, Dr. Gupta, for inviting me to participate in this very relevant, very important discussion and topic. So as you kindly introduced me, my background is in practical implementation of technologies such as blockchain, such as biometrics, specifically facial recognition for digital identity, having designed and implemented a solution for the United Nations Pension Fund to support the proof of existence of 84,000 retirees and beneficiaries of the UN residing in 192 countries. So my initial thought in addressing the challenges that conversational AI and chatbot can have and can present in settings with low resources is indeed, and of course I admit my bias here, is first and foremost the identification of the user. As we are looking at potential use cases, we cannot avoid to appreciate the importance that especially in the healthcare sector, when and if there is a relationship between a patient and a system, using this term in a broad sense, that is intended to provide

Rajendra Pratap Gupta:
services, that is intended to provide supported information, it’s critical that the system has the capability to identify who is the end user. Because as we can imagine, the response will need to be tailored and aligned with the specific needs and expectations of the end user. So here comes the concept of having multiple technologies that working together can create a system and a solution that ultimately is able to address the needs of the end user. So the proposition here is, as we discussed in other panels, is here we have the AI, which is a probabilistic technology, jointly working and functioning with a blockchain, which is a deterministic technology. And the two of them, in conjunction, they can complement each other and provide that level of support to provide and to offer and to confirm certainty about identity, certainty and reliability about the data that ultimately the machine learning model are using to elaborate the responses in the conversation. So I think we can start framing the conversation, the discussion, at least from my point of view, by looking at how the joint functioning of this technology can ultimately create a value and a secure solution for the end user. Thank you. Thank you, Dinu. And I think this is Dinu’s call to everyone that those of us who believe in leveraging AI for health, or for that matter, any critical sector, please ensure that the probabilistic technologies have a denominator of deterministic technology. So AI in isolation is probably going to create more distrust unless you start merging it with blockchain. I think this is where this panel is clearly having top global experts who have done groundbreaking work in terms of trying to get both these technologies working together. And in health, we always say that anything you do, the first thing that the user looks at is with distrust. And when you start saying that this technology has a basis of ensuring an identity and reliability, and that cannot happen without blockchain. And this brings me to my next panelist who is, I think for many years that I have known him, is today the man on the mission, the man who leads AI for WHO and the WHO-ITU collaboration. Beyond that, he knows beyond AI, I mean, given his work in mobile health and other standards negotiating with maybe 194 countries for getting people on board for this emerging technology. Sameer, I want to ask you, what has been your experience? What is your vision? What’s the work you are doing in this area? And what can this conversational AI deliver for the LMICs? Over to you.

Sameer Pujari:
Thank you, Rajendra. And thanks for sitting on this forum. I think it’s a very interesting discussion, especially at two angles. One is the conversational AI because the discussions have floated very much towards just generative AI. I think there are two different components to it that needs to be discussed. And second is the low middle income settings. I mean, that’s the key here that we’ll discuss. Let me step back and say, before I go into my experience part, that often in this technology forums, we focus a lot on technology. I would urge everyone to start, take off that hat and think of people. I think it’s very, very important that any discussions we are doing are focused on the people that they are getting the benefit out of. And these are not one set of people, these are the future generations that we’re looking at. These are the current population that we’re looking at. These are care providers that we’re looking at. Everyone has a role and impact in this area of work with AI and conversational technology. So I think technology has to be understood as an enabler. That’s the first point. Now, second point going in, is it really making an impact? What is the challenge and how are we looking at it? Rajendra, you mentioned at the beginning of the session that there are gaps in health care services. Even in 2023 today, we are still seeing a massive gap in rural Africa where women cannot be screened. It’s very expensive to screen for cervical cancer. We are seeing in Egypt, massive gaps in screening of diabetes population. And these are problems which are not existent because of gap in access to health care, but there is a gap in the health care providers proportionately. So technology provides that specific enabling factor, especially the conversational AI. We are even not there at the stage of population health components, sexual reproductive health areas. We are still missing those massive outreach components with the health care gaps. Forget the part of health education and those things, they are way forward. I think that’s where the role of conversational AI is critical. It has been shown through science that in the low-medium country settings, with the technology getting cheaper, there is a potential that these technologies can make a difference across different disease areas in a very effective manner. And our Director General mentioned that very specifically in the July launch of the Global Initiative on AI for Health. However, we have to be very, very careful of four areas. First is equity. And I think that’s where the main component comes into play is when you’re trying to deploy technology in low-medium country settings, the business value for this is much less. And hence, it is us as this forum or the civil society groups or the international development groups who need to be cognizant that we are working in ensuring equity. The technology companies are not going to be pushing for that. I think that’s one part that everyone has to focus on as we look forward. The second part is collaboration. It’s extremely important that we work together across sectors for health, education, and different areas and domains of work. Third is focus on sustainable business models. It’s very exciting to trigger a new product, a new project, and go to the field. And 90% of the times, I’ve seen it die because it doesn’t have a sustainable business model. So that’s a very important component. The fourth point is looking at how it benefits the user at the end of the table. That’s the most important. How can you take this AI to the people? That’s the discussion that has to be happening all throughout. I think if we can focus on this around the people-centered approach, the people focus, we can make an optimal impact with conversational AI in the changing of the healthcare domain across the SDGs indicators, not just the healthcare indicator, but across SDGs. And that’s the key, I think, I would like to see this forum and the members of this in their own role focus on. We have what you asked me earlier, what we’re doing, WHO together with ITU and the World Intellectual Property Organization, heads of the three agencies, announced a global initiative on AI to bring together all this work to kill the verticalization or to reduce the verticalization so that we can work on WHO side providing the guidance, standards, policies on the facilitation side, bringing all the groups together, and then actually helping countries implement that member states level through the right governance approaches that we have. So I think at this stage, I’ll summarize by saying that there’s a huge potential in the healthcare market that our member states are seeing, and they’re asking WHO to work towards that direction. However, as the UN body, we need to ensure that our member states do not get hit by the private sector business models, but we can benefit everyone in the private sector in terms of maximum value of AI in healthcare. Thank you. Back to you.

Rajendra Pratap Gupta:
Thanks, Sameer, and really heartening to hear what most of the people refrain from saying is that we should not get trapped by private sector business models. At the same time, we have some phenomenal people in private sector, and I think to your point of people benefiting, we have next Mr. Sabin Dima, who is the founder and CEO of humans.ai. I really like his approach and the way he’s building the AI verse is that he says you will be able to do anything you can think about with AI. I mean, it is totally disruptive approach that Sabin has. Sabin, it’s nice to have you. I know you are traveling. I would like to hear your views given the work, and I would like you to maybe speak for a minute about the work that you have done in this field and how you are disrupting and what do you see as the role of conversational AI in healthcare. Over to you.

Sabin Dima:
Hello, everyone. Thank you so much for this opportunity. I agree with Mr. Sameer that we need to have a human driven approach. I’m Sabin Dima. I’m the CEO and founder of humans.ai. We are for more than four years in the AI field. In this new AI era, we are a company that already have some world premieres. We created the first AI counselor of a government. We created an AI that it’s able to have a conversation in real time with 19 million of Romanians, and all of those opinions, we’re using them to train an AI. And on the other side, the decision makers can have a conversation with one entity, with one AI, like they’re talking with 19 million of Romanians. We strongly believe that artificial intelligence is the greatest tool ever created, and in order to democratize it, we created an AI framework that makes it so easy to create a narrow AI. I don’t think that AI is able to replace humans, but it’s able to replace some skills. And we can help with that, but we’re taking in consideration two major aspects. One is the data, and for that, we created the first blockchain of artificial intelligence in order to have the data traceability, to create what is called explainable AI, and to make sure that if I’m giving an opinion to this governmental AI, the AI will be trained with my opinion as well. And there is no bad actor that can delete that opinion. And the other one, it’s ethics. We had a lot of research papers on ethics in AI. The latest was presented at the Imperial College in London. Regarding the AI in healthcare, for sure AI is going to democratize access to health, but I see a bidirectional approach. Usually, we are using conversational AI to get answers. We are interacting with the AI. We are asking questions to the AI. But everybody wants to solve people’s problems, but I think we are not aware about those problems. So we should engage in a conversation with people in a human-like interaction like we’re having right now, to understand what are the people’s problems, and probably the most important, what is the sense of urgency? And for that, we’re not asking governments to invest in infrastructure. We’re not asking for governments to invest in hospitals and so on. We need just an internet connection. And in some cases, there are some AI models so efficient that can be encapsulated in a blow-entry tablet, and we can ship it in remote places. So I think that we should build this bidirectional conversation to ask people to be aware of what are their problems, and on the other side, to encapsulate different doctor skills, being able to respond and to create this open innovation platform, that it’s a living organism, that any startups can participate and can bring different skills under the same core.

Rajendra Pratap Gupta:
Thank you, Sabin. Coming to Mewish, Mewish, you have been in this field, writing policies and roadmap for telemedicine, and now this new field of conversational AI. And given the fact that you are involved in academia, which is expected to show the roadmap for the future to those who are into this field, what’s your view about conversational AI?

Mevish Vaishnav:
Thank you, Dr. Gupta. Hello, everyone. I’m Mevish Vaishnav, the Group Chief Operating Officer at Digital Health Associates and Academy of Digital Health Sciences. I thank the DC Digital Health for this extremely important session. While we talk about generative AI and large language models like LLMs, I would say that the basic and the disruptive point for LLMs and for generative AI would be the conversational AI. Just imagine if we had a conversation a scene where billion people are speaking and billions of people talking to patients, populations, speaking about health issues, and clinicians addressing them. It would make a phenomenal opportunity to analyze these conversations and create the DHAI, that is the generative health artificial intelligence, which would be different from artificial intelligence for general purposes because health is very technical. It is clinical. So, I see a great opportunity of conversational AI being the starting point for the generative health AI, which will over time kind of eliminate the need for using doctors for basic health problems because most of the people in the rural settings or in the semi-areas or even urban areas have the need for basic information and this can be handled by conversational AI, which is driven by either generative health AI, but both are dependent on each other. Without this data, we actually cannot do anything. So, I see a phenomenal opportunity and I think we should build upon this. Thank you.

Rajendra Pratap Gupta:
Thanks, Mavish. Moving from what you said is the generative health AI and the fact that when people start interacting over voice, over communication, rather than texting or writing, which limits their ability as, I’ll not call it illiterate, but digitally illiterate populations who have not actually learned to write still. I mean, that’s a major part of the population. In fact, yesterday at a panel, we were talking that 2.6 billion people are still not connected to the internet and with what Sabin was saying, you know, of shipping the tablets to low resource settings. I mean, just imagine if people start talking, you know, the quantum of data that comes out is going to be exponentially more than what we have today because today you have to type, you have to text, that gets captured for analysis. The moment you start analyzing voice-based data is going to be exponential than what we have. So, I think the accuracy will increase and that will become much more worthwhile. But I think at this point, I would like to bring Ashish Atreja, the actual person who has done a lot of work in this during COVID and even earlier. Ashish, what has been your work in this field and how do you see this field shaping up and the role of conversational AI? Over to you.

Ashish Atreja:
Dr. Gupta, it’s a pleasure to be here and thanks for having me. Greetings from California. It’s 1 a.m. here and really excited about this. We just launched the largest network in the United States on generative AI in healthcare called Valid AI. And the reason being, just a very brief background about me, I did my medical school from India and then came to the U.S. to do public health and then informatics. So, as a physician, I’ve been practicing for the first 10 years of my life and now as an informaticist technologist supporting technologies for the University of California Health and now working globally in many things. I still am an adjunct professor in medical school in India. So, I’m considered an app doctor because I started building apps around 15 years ago and these were mostly deterministic models. So, we took the rules from the guidelines and the biggest gap we see is, and very eloquently expressed by the previous speakers, there is efficacy which we see from the medicine, what is possible today, like 99% of patients’ blood pressure can be controlled with the current medicine. But the real gap, which Dr. Gupta mentioned, is 80-80-80. Many patients don’t even get access to the doctor. They can’t even drive to a doctor. The doctors have a waiting list and even if the doctor prescribes a medicine, they do not have time to explain how to take it and how to do other things like salt reduction and others. So, there’s a biggest difference in the care which actually patients get in their home and then what is possible. And that is because most of the human medical care globally, whether it’s United States or Africa or India, is because we have locked medicine and care into the same time and the same space as a physician. So, everything has become physician centered. You have to come into the same clinic or a hospital to get care. What generative AI and AI can do finally is unlock care with the time and space. So, you can provide care anywhere you can and you do not need a physician. You can extend beyond one-to-one physician-centered care to what we call as exponential one-to-many care. If I have to tell the same thing about blood pressure control, I can make myself into a conversational AI bot. Now, with generative AI, within a matter of weeks and I can deliver not only to people which I see in University of California, I can deliver across California, across US, but really I can now deliver across globally. Right? So, any solution now we can make, if that is validated the right way, can immediately become a global solution. So, we are finally at the cusp of unlocking the biggest supply demand issue in healthcare by democratizing it completely. And if you really combine the rule-based stuff, the guidelines to provide rule-based care through text-based, you can then combine that with generative probabilistic. You’re unleashing the science of rule-based care with the conversation which patients need. Because rule-based care is our scientific way of physicians doing it, but conversation is the way how patients get it. And that has always been a barrier how to bridge it, but now with combination of these two technologies, we call it a hybrid AI, you can combine the physician-centered care traditionally with patient-centered care which everyone needs today globally. So, really excited about this. We have all the US states now looking at, you know, and really we need to go with a problem-centered approach first and really looking at equity. The equity is not just in, inequity is not only in patients, inequity is in countries, in states, and in healthcare organizations. If we do it the right way through collaboration, which is really I’m looking for here, we can finally make it the most inclusive, the most democratic way of providing care globally and become, go from digital divide to digital bridge. And I think that onus is on us, not on technology. We humans are the transformation agent to bridge the gap and really it’s a big calling for us to really leverage technology, but put our own DNA and purpose to bridge the gap.

Rajendra Pratap Gupta:
Thanks, Ashish. And this is very important of the fact that you launched valid.ai, I think, at Health in Vegas, I guess it was, as we are here, I think it’s going on parallelly. I move to the next expert panelist, Dr. Olabisi. She is a pediatric doctor and she has done phenomenal work using WhatsApp and others in underserved populations. I mean, coming as a clinician like you, Ashish, she has done phenomenal work. So Dr. Olabisi would like to hear about your work and what’s your suggestions. And I think you have a presentation, so I’ll ask the technical team here to allow you to just share your slides briefly. Hello. Hello. We can hear you, Dr. Olabisi. Please go ahead. I want to thank you very much, Professor Gupta, for inviting me to this forum. So I’m a pediatrician.

Olabisi Ogunbase:
Okay. Greetings to everybody. I work in a general hospital, a maternal and child center, and we see children, you know, mothers bring them to the hospital and they go. So we have no contact with them thereafter. So we thought of how do we continue and ensure patient engagement? You know, how do we ensure that we still maintain, you know, a form of interaction when our patients leave us, you know, so that we can prevent relapses and what have you. So that was what brought us to, you know, thinking of what to do when patients leave us. And that’s how we came about digital technology and what to do when they leave us, how to involve them in their own care. So we thought of using WhatsApp as the means by which we communicate with our patients. Please give me a minute. Okay. So we thought of using WhatsApp in communicating with our patients and, you know, having that relationship with our patients, even when they leave us. So I’ll be taking my presentation in this outline, a brief introduction, definition, objectives, and what actually we do, advantages and no. So WhatsApp is a form of digital technology where we use tools to maintain that relationship and that engagement with our patients. So for us, mobile phones is what we have and mobile phones is what the patients also have. So that’s the tool of digital technology we’re using. So patient engagement is how we are involving patients in their own care and digital means we’re using an electronic means to ensure that. So when we started our objective was, okay, how can we pass information across to our patients? How can we pass notices of what’s going on the hospital to them? How can we educate them beyond the little time? You can imagine in developing countries, there’s so many patients. So you don’t have that much time engaging with them when they come. So what’s all that means? Can we use to pass education to the patient? And the forum also served as a support system because the mothers engage amongst themselves on that WhatsApp platform. And they support one another, they ask questions, they share ideas. And those times we just stay as a fly on the wall, we don’t say anything. But when they now ask us questions, we can now come in and answer the questions. So there are many advantages to this form of engagement, digital engagement using WhatsApp. For us, it’s, of course, optimizing efficiency and unnecessary visits to the hospital. I mean, we can answer some questions, they don’t need to come to the hospital. And so it improves quality of life, it improves patient safety, it improves health outcomes, because we’re still engaging with them, we still have that contact and relationship with them. So there are many advantages. So as of a few days ago, you can see here, there are about 395 participants. And this is just one of the WhatsApp platform. Every clinic has a WhatsApp platform, a dedicated WhatsApp platform. So from the picture, we talk about weight gain and their weight increasing. The mother sent pictures to us about different things about their children. This is one saying the hair on my baby’s head, you know, has gone off, what is happening? There’s on the Leica, you know, what’s happening to my baby’s cord. So they send pictures. So they type questions, they can send pictures, sometimes they even send voice notes. Then doctor, listen to my baby’s breathing. I’m not comfortable, just take it. They record the baby’s breathing and they send it to us. So we are also able to listen to the breathing, we’re able to read their text messages, we’re able to see the pictures they send. So these are various from them. This here is showing the information. Sometimes it’s World Breastfeeding Week, it’s World Pediatric Day, it’s World Hand Hygiene Day. We’re using that forum to educate the mothers on the platform. So these are just examples of, you know, interactions that I picked up from the WhatsApp platform. They ask about immunizations, or they ask my baby has cough. Doctor, what do I do? And I see that we’re able to interact with them. Okay, come to the hospital, do this first aid, let me see you tomorrow. So we’re able to book appointments, we’re able to see them. So we’re able to interact with them. So that aids the experience that they have. So this is pictures we also send to them. This fontanel is normal. This is how you engage better when you’re positioned, better when you’re breastfeeding. And this picture on the bottom right is a picture of the rash. So they send us, doctor, look at what’s on my baby’s skin. What is that? Do I come to the hospital? What do I use? Of course, we don’t really prescribe on the platform, but we can educate, we can inform, we can say, okay, I need to see you in the hospital. come at 10 o’clock, please come at 9 o’clock. So it’s a forum. And these are pictures of their babies that they send on the platform. When their babies are six months, they say, this is my baby. I’ve completed exclusive breastfeeding. They’re excited because we’ve talked about exclusive breastfeeding. So they send their babies pictures. Like I said to you before, they support one another. So when a baby is one year, a baby is six months, they send a picture. All the mothers are congratulating them. Oh, you’ve done well. You’ve breastfed exclusively. I know that we all know here that we’re talking about digital health. We’re talking about breastfeeding is one of the childhood survival strategies. So it’s a big thing for us. So in conclusion, I’ve talked about how at Matana Child MCC at Lagos, we’ve used the WhatsApp platform as one of the digital tools to engage with our patients, even when they have left the hospital. The consultation shouldn’t stop in the doctor’s office. Like the last speaker said, it should still continue beyond the doctor so that we can prevent relapses, we can continue to educate, and all that. So the key words, in conclusion, the key words are digital patient engagement, digital technology, mobile health, using the smartphones that the doctors have, the dieticians have, the nurses have. And this platform is not doctors. Everybody’s there. The nurses are there, the dieticians are there, the social worker. So if the question comes that concerns the nurse, she answers. If it concerns the pediatrician, I answer. Everybody’s on that platform. And it’s really a useful platform for us. So thank you very much for listening.

Rajendra Pratap Gupta:
Thank you, Dr. Olabesi. And I think this convinces us that if you can use WhatsApp to bring such a change, and you get photographs from your mothers who say that this is what the check-in looks like after six months, I think one of the things that you pointed that we don’t prescribe over WhatsApp. But I think what my friend Dinu, who is sitting on my right, has working on technology with blockchain and what Sabin is doing, I think the moment we are able to put the identity within the system, I think the day is not far when I think a perception on WhatsApp may be legal as well. I think that’s the day we should look forward. But I think seeing your presentation, your work that you have done, I think low-resource settings are the high-opportunity settings for conversational AI. I mean, that’s what I would say. And this brings me to my next panelist, Shauna Hoffman. Shauna had led global roles at IBM, Watson, and before that with Dell. She was revered in this field. And she is doing path-breaking work in terms of what she does at GodRate Technologies. Shauna, over to you for what conversational AI can do and what you would say in terms of re-fencing the negatives around conversational AI. Over to you.

Shawnna Hoffman :
Thank you so much, Dr. Gupta, for having me here today. And Dinu, I love sharing the stage with you. There are so many great insights that you have. And I’ve been in artificial intelligence for almost 20 years now. And I have seen it at its best, and I’ve also seen it at its worst. And when Watson won Jeopardy back in 2011, I knew that conversational AI had actually taken the front seat. And so I joined IBM at that time. I led a Watson practice for Watson Legal. And when COVID-19 hit, I was chosen as one of the few to lead our COVID-19 solutions to the marketplace. And we had three. And one of the things I had realized after leading an AI practice, that AI wasn’t enough. And we needed to be responsible. And that responsibility was tracked and traced through blockchain. And so that was the combination of both. Our three products that we brought out to the market within the first three weeks of the shutdown were one of really important. Remember when we couldn’t find masks and we couldn’t find gloves? And it was really a challenge to get the PPE across the globe. We had an AI chatbot combined solution with blockchain to track and trace all of the materials. We found over 10 billion within the first 24 hours. And so it was connecting people all over the globe. I will say that AI, one of the most amazing things for healthcare is that those individuals who can’t often travel to a location to get to a hospital or get to a doctor, often they have mobile phones. And so conversational AI is extremely important for us to get around the globe so that individuals have an opportunity to get forms of healthcare. And maybe it’s unusual, it’s not traditional, but it answers those problems, as our previous guest speaker just said. And I love what you’re doing to really bring that, especially to mothers. I’ve got three kids of my own, and man, did I have a lot of questions when they were little. Because every little cough makes you a little scared as to what’s happened with them. So other solutions that we’ve worked on, of course, the supply chain. And making sure that not only, oddly enough to say, not that doctors are a supply, but during COVID-19, they were really lacking in so many of the areas. And so we were able to move doctors around through, again, our chat bot. The doctors were able to chat to say, hey, we’re available, we’re happy to go anywhere in the globe, and we could connect them with the hospitals that were the most in need. Again, a blockchain solution with AI. You know, conversational AI has such a potential to bridge the healthcare gap. And I would definitely say there are five that we have worked on throughout the years. And I have to say this before I even mention the five. AI has been around since 1956. And the newest, most excitement that I’ve ever seen is really just this past year. And it is when a system that used to cost my clients over $20 million to put in place, that was Watson, is now a conversational AI that is free to the globe. And so we’re seeing a lot of hype, a lot of excitement around there. But do know there’s a lot of use cases over the past 15 years that IBM Watson has been around that they’ve really solved a lot of these problems. And so there is a good company to go back to to ask those questions. I don’t work for them anymore, but there are a lot of us who have that are willing and very willing to share our experiences. If we were to look at the fives, let me jump into those, accessibility. That was mentioned by our previous speakers. Reaching the remote and underserved populations that lack that access to traditional healthcare. Again, access to mobile phones. Many of them have, although we did talk about earlier, yesterday, the two points, and you did here too, the 2.6 billion people that don’t have access to the internet. We need to fix that to be able to give them an opportunity to be part of this global health system. I love the consistency of AI, so 24-7 availability is my second one. It’s extremely important to be able to have doctors, which we’ve done in the past. So Watson had an, we did a lot, even remote surgeries. That kind of gets into robotics. Again, AI is over 90 different components. Conversational AI is only one of them. You can do remote surgeries from one end of the world to the other, and so we had some really amazing things that we saw. But again, that 24-7 availability with conversational AI is extremely important, and it is consistent. I will say, so I’m the president of Guardrail Technologies. One of the reasons that we exist is to put guardrails around AI. AI, as Dino had mentioned, is a probabilistic model. It is not correct 100% of the time. Sometimes it’s even really incorrect. We’ve been working in the medical space in AI. I’ve worked a lot with various different hospital systems in the U.S. I just spoke at one about six weeks ago, and we dove in with 30 of their top physicians to figure out what we needed to do to answer the problem of the AI being wrong and the AI hallucinating. And it could be very scary that it gives the wrong information and could actually cause death. So we need to be careful. We have guardrails. We fact-check the generative AI. That’s part of our program. But make sure that you are fact-checking it because it is going to be incorrect. The best systems out there, because it’s probabilistic, none are going to be 100% correct because of the way the model is, and there’s nothing wrong with that, but we just need to make sure we’re adding that extra layer that confirms that we are fact-checking our information. Education, great educational tool, making sure, as you had seen, that the mothers know how to breastfeed their babies, what the different rashes look like. I love this one. Language and cultural sensitivity is one of my top five because AI can be used to be customized to the local language, the local responses to things. It can be really cool. There’s some AI out there. I just was talking to one of our previous guests, and he had mentioned that they have a movement program that he was in the midst of going and finishing up, working on a patent application for. But as an individual moves, the AI can watch the movement and see what possible types of medical issues that the individual has. There’s some really good language of cultural sensitivity, but then also from there, being able to take that and say, okay, that’s a cultural thing, but then this is just unusual, unique. They may have symptoms for other things. Again, very customizable to the individual. And then my last one, efficient triage, which we can identify urgent medical issues, not, again, 24-7, not having to wait for a doctor’s office to open. So thank you.

Rajendra Pratap Gupta:
Thanks, Shauna. This makes it very interesting to first see those who have been into the clinical side have used it at scale. So there’s no doubt about the effectiveness. In fact, it’s about saving lives. What I said in the beginning, the three ATs, the 80% have no access, 80% can’t afford, but 80% have acute problems. That means they every time don’t need to go to doctor. And these are 80 As, all As, access, affordability, and acute. So the fourth A would be artificial intelligence, of course. But given the fact that we are DC Digital Health and we believe in tangible outcomes for what we discuss here, and that we have taken a topic of conversational AI, so I’ll go back to my expert panelists and ask them that if you had a clean slate and given the discussion that we have with our expert panelists, what would you recommend, Dinu, to you in terms of our pathway for the next one year for this field? Thank you.

Dino Cataldo Dell’Accio:
Thank you very much for that question and also for that call to action. So I think the previous sharing and comment were extremely relevant. I really like the observation on human centricity, the distinction that was made about the gaps, how to bridge the digital divide, the concept of guardrails that Sharna just mentioned. And so here again, I like to talk from personal experience. I’ve been working for the United Nations for 22 years. And my background is actually in auditing. For large part of my career, I was the Chief IT Auditor at the United Nations before becoming the CIO of the UN Pension Fund. So I have a professional deformation on assurance, on evidence. And I think that one of the implicit concept that if I may, all the speakers of this panel have touched upon, but we have not yet made it explicit, is the concept of trust. In order to have attention to human centricity, in order to bridge the gap, in order to enable the human being to approach, to make use of, to be supported by these technologies, I think we need to also build a framework of trust so that they don’t need to understand what is the distinction between a conversational versus a generative AI. They don’t need to understand the distinction between a blockchain or distributed ledger technology. They don’t need to be bothered with those technological details that are often too complicated to explain and to verbalize. They need to just be able to trust the solution and the entities, whether they are private, whether they are public, that are offering the solution. So I believe that it’s incumbent upon us working in this field to come together and start building bottom up, and of course also top down, a framework of generally accepted criterias and principles that can be utilized to then support the reliability, the trustworthiness of the solution and this technology where and if they are indeed implemented in healthcare or for that matter, also another area of our society. So I think that there is that need to start now looking at the fact that I think as we all recognize, this powerful technology can be used for good or for bad, and not all solution have the same level of reliability. So there is a need to start having some sort of criteria that will enable us to start comparing and contrasting, to make assessment and to then providing a level of assurance that I think that ultimately that the human centric approach calls for the user to deserve that what they are going to use is trustworthy and is reliable.

Rajendra Pratap Gupta:
Thank you for raising this very important point of addressing this core issue of the human centricity plus reliability. I think it is a twin opportunity and a twin challenge too. And this brings me to Sameer. Sameer, you lead all the AI initiatives at WHO and WHO is the multilateral body where every government looks up to. I think now there is an excitement across the world for generative AI and AI for health. Everyone is waking up. So what is your advice and the roadmap for the next one year or the action plan, if I were to call it?

Sameer Pujari:
Thank you, Rajendra. And you rightly said all the member states are actually getting very excited about this work, both from a positive side and negative side. And I said negative excitement is a scare or the fear of what damages can do, and the positive excitement is the opportunity. So I think there is, and we see an unprecedented push from member states. Normally, there will be a two-way discussion, but this time member states are actually coming to us and asking, and not just now, since December last year, when actually, charity started picking up the speed. And that’s when it started off. So WHO actually, through this process, has put out a position in the June WHO Bulletin, where we have clearly articulated the value possibilities of generative and discussional AI. And in summary, the one sentence that summarizes that article for everyone here is, is the British position is to be cautiously optimistic and apply right safeguards. I mean, that’s what we are saying here is, we have to be cautious, we have to be optimistic. And as long as we have the right safeguards, and when I say safeguards, it is the ethics approaches. And now, ethics is a very common word. I mean, it’s a moral word, almost to put out and discuss, but I think it’s the application of ethical use, development, and deployment of technology or AI specifically. More importantly, because AI has more power than before, is a critical part. And WHO has guidance, which it’s working with its countries to deploy. So it needs to be not a knee-jerk reaction, but a more sustained governance approach for AI, because AI is here to stay. It’s already with us. It will make a difference in the way things are going forward in terms of healthcare, in terms of development, in general, in terms of education, agriculture. So I think what’s important is, we need to take a detailed, systematic, creative approach in this. in these regards. Also, regulations. I mean, we don’t want to have regulations again becomes the whip for our times for the developers. But what we want to make sure is that regulations are there to safeguard and provide guardrails for the right technology and the right products to be deployed across the domain. And I think this is also humbling this time to see that it’s not just coming from the countries, it’s also coming from the developers, the industry, the private sector. The recent discussions at the US Senate through all the CEOs, where there’s a call for regulating AI through the governments and the UN. I was in Copenhagen just last week where there was discussion of the UN High Level Commission of programs on AI and regulations and governance. And there’s a huge push from the Secretary General on sort of putting this work together. So I think that’s the area where the world is going towards. And we will need to prioritize that. Again, keeping in mind that it has to be people-centered, not technology-centered. So the regulations, the ethics should not be technology-controlled or centered, but people-centered. How is it going to make an impact? And they have to be adaptable for different countries. As you mentioned, 194 countries. There are different stages. And again, for the first time, we’re seeing a rather less gap in terms of preparedness between the high-income and the low-middle-income country settings. I mean, there is some, probably, parity there. But there is still a similarity across the board. So I think it’s important to manage that. Use the power of collaboration. And I think that’s what we’re leveraging through WHO, through ITU’s work. And I think the forum there, a lot of the colleagues from ITU are there to leverage what is existing. As WHO, we’re getting normative guidance, which is science-based, evidence-based, and deploying that science-based ethical regulatory approaches. And my call for this community here, which has a mix of a lot of expertise, many grassroots workers we’re working on, to ensure that the guidance that has been deployed, or the products that are being used, are not technology-centered. It has to be science-centered. And when I say that, it’s the guidance, which is the content that is coming into it, should not be written by the developers. And this happens. Developers pull out Google, take the content. They are more fancy towards the application part of it, not on the content. But I think, as actual healthcare providers, our job is to make sure that the content is governed through the right full mechanisms, the process is right, we’re done. And technology is the enabler, which is a massive, massive boon for the healthcare process. And if we can do that combination right, in the ethical and regulated fashion, pushing towards the right governance mechanism, I think we will have a successful one year of AI. And I hope we can come back next year in this forum and say, we just talked about it in 2023, but 2024, we are making impact. Back to you, Rajen.

Rajendra Pratap Gupta:
Thank you, Sameer. I think in this IGF annual forum, the 18th forum that’s going on, we heard, we had the entire high-level panel that was constituted by the UN Secretary General at this meeting discussing AI. I think one of the things that I saw in this forum is that most of the sessions this time are on AI and native AI, and around various guidelines. And you made this point of consciously optimistic, and also about regulations. But in a technology which is evolving, how can you regulate? I mean, do you think that it is AI that will itself regulate or there’ll be just guidelines which people should follow? And I think the work that Ashish, what Mavesh and others are doing, will that be a good starting point? Because guideline gives you a general direction and doesn’t stall innovation, because regulation to a point can become like a hindrance to innovation. So, I mean, do you think that we should stick to guidelines rather than regulation for now, for next two, three years?

Sameer Pujari:
It’s a great question, Narendra. And I think there’s no blanket answer to it. I think even the European Commission, EU Act is looking at segregating the different ways that we regulate products. And so I think it depends a lot on the solution, on what kind of solution we’re talking about, and what is the impact of that solution to define whether we can work with guidelines or regulations are needed to do that. And I think it’s very centric. So let me give you an example. For tobacco control or diabetes prevention and management, it is a prevention. There is a lot of content which is available. These are healthcare programs which have provided guidance, which don’t have the outreach. I think such simple guidelines, guidance-driven programs for health education, personally, I think can be very quickly distributed if there’s a small mechanism of testing the right content in there. There is a risk on the hindsight that if you don’t control or regulate these content, this can damage by providing misinformation in that, which is a big concern. So there are some products which I think can be loosely regulated, guidelines-driven, but there are some specific areas where cancer screening products, where there are more diabetic retinopathy screening programs where it needs to be regulated in a way. Now, I get this dialogue all the time in discussion whether regulations is over-controlling innovation. And I think that’s the thin line where we have to draw, how does the value? I think the member states or the countries want to use the value. They have seen the problems that they have and how technology can help that. So I think the intentions this time are around more focused on how can we maximize the value of technology, but at the same time, having that regulation is important because without regulation, there’s a massive risk of misappropriation, misuse. So I think the regulation, the level of regulation and the control of regulation needs to be properly adapted and uncharted and defined, but it is important to make sure that we are not transversing ourselves into an open sort of platform where anyone can do anything around healthcare, especially in healthcare. And I ask this question to people, would you say the same thing around when it comes to your financing? Would you allow non-regulated financial digital models to work across the board? And would you be open to doing that? Health is two domains further. So people are more worried about the money than the health, unfortunately. And that’s where the answer comes in. I think it is important to be able to regulate rightfully so we can benefit the value of the technology opportunities at the same time, control or safeguard the damage it could cause in the long run.

Rajendra Pratap Gupta:
But Sameer, even after the Sarbanes-Oxley Act in the financial markets, we had the subprime crisis. We had banks collapse even a few years back, even this year, Silicon Valley bank collapse. So I think even if you over-regulate, we have the outcomes. I mean, wherever there’s money involved, there would be, I think you made this very interesting point in the very beginning, and I really appreciate you for that forthrightness is to not get into the trap of the private sector. So, but I think the experience of over-regulation hasn’t served the purpose. I mean, that has kind of been the government’s way of putting it, saying we are pro-people, so we need to safeguard them, but neither the safeguard the people nor the organization. At the end of the day, the sector bleeds. But I take your point. Coming to the fact of point that you said about people-centricity or people having trust and ethics about it, so I would go to Sabin who is actually building at world scale. Sabin, what I understand from my experience of leading consumer-facing organization is that trust is a matter of value. So, if I get value out of humans.ai, I would love it. If I get value and benefit out of the products and services you roll out in AI, generative AI, anything, that will create trust. So, what would you think about the conversational AI product or healthcare in general in terms of using AI to create that value for creating that trust? Because value is the precursor for trust, not the other way around.

Sabin Dima:
I’m 100% sure that the technology is here. So, it’s not a problem of technology anymore. Even us on this round table, we have all the resources to start experimenting with the project. I believe a lot in learning by doing. I believe that if we will have together as a group, one use case, we will help the regulators to better understand and we can fill this gap between real world and the regulatation of the area. So, I will choose the easiest win that we can get, probably in the aftercare. For example, we have a project with a big pharma company. We saw that in our region, it’s a huge dropout rate. So, people are not finishing their treatments. After three days, they are not finishing their treatments. After three days, they’re feeling better and they stop taking their antibiotics. So, what we are doing, we are cloning the doctor voices because the doctor is the only authority in your life when you’re speaking about medical treatments. So, we are sending audio messages on WhatsApp with the voice of the doctor saying, hey, Sabin, I know that it’s day three and you’re feeling better, but it’s important to finish the… So, what I’m saying is, if we together, we will implement only one solution and we will choose one region, we will learn a lot and we will learn from the real world what were our initial ideas and what the real use case outputs really look like. So, I’m willing to help with our technology and our team and our expertise to create together a real-life use case in conversational AI for healthcare. And in one year from now, we will know more that we know now.

Rajendra Pratap Gupta:
And Sabin, would I take the liberty to say your message on your behalf, what you always say is, for AI start now, do something rather than just thinking about it.

Sabin Dima:
Exactly. The technology is here, we have all the skills. I see a lot of passionate people about the subject. So, we need to start doing.

Rajendra Pratap Gupta:
That’s the best message. And I really remember the line that you said last time that when you heard yourself speak in Portuguese, you were able to actually check what phenomenal opportunities exist before us and the project that you’re doing where you clone doctor’s voice and convince the patients to carry on with the treatment. So, I think the fact is that conversational AI has multiple use cases. And so, one of the things I understand, and we carefully picked up this panel, it was not because of friendship. It was because of the complimentary things that come to the table by thinkers and doers and regulators who are critical to success of conversational AI. So, that’s why we have blockchain, we have AI, we have WHO, we have UC Davis, we have Academy of Digital Health Sciences, we have Shawna. So, this is what is the beauty of this panel is that we should be able to get into something decisive which we can measure over the next one year. Mavish, coming to you, fact that you run a couple of initiatives in digital health, what would your action plan be for the next year?

Mevish Vaishnav:
I believe conversational AI can actually serve as a powerful tool in patient engagement, educating people about the facts behind a particular health-related issue. As you rightly said, Dr. Gupta, imagine the effort that would go in typing and texting, but conversing would actually leave an important and exponential impact. We all know the time that doctor spends with patient is very less, but if we have a conversational AI, patients would be happy that they have been heard. And at Academy of Digital Health Sciences, we are working on the report on generative health intelligence, and we will be releasing it soon, covering all these topics on the role generative health intelligence will play in shaping the future of healthcare intelligence. We will be happy to collaborate with you all, and I would also like to say that within the dynamic coalition of digital health, Dr. Gupta, you should take the stewardship in creating the global group because UN is the largest multi-stakeholder and a multilateral body, and getting everyone under the roof to form a global generative health AI group and leaders where you have regulators, policy makers come together to give a direction to all stakeholders, doctors, hospitals, and the frontline health workers to understand how generative AI work, how to get trained on it, and how to deploy it. We already have a course on digital health at Digital Health Academy, and you can visit the website to know more about it. Thank you.

Rajendra Pratap Gupta:
Thanks, Mavesh. Ashish, over to you after your grand initiative that you launched this week. What are the opportunities for stakeholders to work together? Because the worst thing that happens to health is we all keep doing our work in silos. We rarely connect, forget, hear, and listen, and come together to act. I mean, when I took over as chairman of the Dynamic Coalition on Digital Health at the UNCGF, one of the things I have done over this last one year is to get all the people in the same, I would say, wavelength, pick up a project, and deliver it. So every year, for all the Dynamic Coalitions that at least I chair, we come with tangible outcomes every year. So given your leadership and your pioneering work, what do you suggest we should be doing in the next one year? We heard your previous experts from the position of authority and influence.

Ashish Atreja:
Happy to. I think one of the critical things is I think it’s the onus is on us. There is a very famous map called Gartner Hype Cycle, where it shows about all the technology that comes. There is the hype peak that happens. Then there is a valley of disillusionment, a valley of death. And then there’s a second wave, which comes later on. And generative AI is now at the peak of that hype right now. But we all know there’s a value. So where I would echo is the transformation peak, that is a second peak, that is slower, that happens after the valley of death, is the true peak. And that is one, as humans, we don’t just look at technology, what it can do, but actually we start learning how to use the technology for the right use cases within our workflows in a trustworthy, scalable, scientific manner. So it’s repeatable, replicable, right? And that’s what science role is, right? It takes what one person may say, but actually validates that approach across multiple different variations. So you can be fairly confident, for example, if I give this blood pressure medicine, this is gonna be the impact on it, because it’s been repeated replicable success. So we need to go the same thing with AI, we need to have that lens, similar to what Sameer mentioned, put that scientific evidence-based lens. And then see if something Sabine is doing great, can we replicate that across country? And can we demystify that through a playbook? We call it an implementation science playbook. So through valid AI, 30 health systems, health plans have got together. We have three global partners right now, in Israel, India, as well as in Canada. But our goal working, I love the suggestion which Mavish mentioned is, creating this global thought leadership group on generative AI in healthcare. We love to contribute our collective knowledge from US through valid and coalition of healthcare AI into it. So we can all learn from each other faster. We can also support each other best practices. And also maybe the ecosystem not only just had to be scientific, but also equal input from our key ecosystem partners, including startups, bigger technology, pharma. So we hear from them. So if we have to do a balanced approach, we don’t err on the side of caution necessarily, but err on the side of optimism combined with caution and have feedback from all quadrants.

Rajendra Pratap Gupta:
I totally agree with you, Ashish. I think it’s a great approach to make sure that the excitement is also backed by competence. And for that, everyone needs to work together. And I think Samir, Sabin, Mavish, and Dino has very carefully told that not only these are the challenges, but there are also technical solutions, which are there. And I think the line that Dino put it, which sums up the challenge plus opportunity, probabilistic plus deterministic, as simple as that. And both the solutions exist at scale. I mean, he is sitting where he has deployed, how many countries is this Dino? What do you have done for the pensioners? 192. 192 countries. We have Samir Pujari sitting here, 194 countries. We have Ashish Atreja, 15 systems. Mavish running a course globally. Dr. Olavesi, I’m going to come to her next. You have everything on this current screen, where everyone who is an influencer. at large and a do or both, which is a rare combination. And we know, I think the Gartner’s hype cycle, see sometimes those historical rules and equations also get challenged. We should challenge the Gartner hype cycle and we should actually make it hope and heal cycle. You know, there is a hope, let’s use it for healing. I mean, as simple as that. So Dr. Olavesi, you have heard all those people. You have used technologies and was very impressed to see this six month pictures of the babies. So given what you have heard, what do you need sitting there in Lagos, you know, from people on the screen to take your work to the next level? What should we be doing? What you should be doing? Over to you.

Olabisi Ogunbase:
Thank you very much for that question. With this WhatsApp platform that we have with the mothers, I can see lots of gaps because when the mothers send the pictures or type their questions, it’s not real time. I might not see it at that point in time, but conversational AI, it’s real time, you know, and it’s all the true machine learning, the responses that are appropriate, that are relevant, comes to the patients immediately. Unlike me, it might be hours before I listen to that. So I can see the advantage of what we are doing, but I can also see a lot of gaps. And it’s not personalized, it’s open to everybody, the 300 and so patients on that platform. It’s not personalized, it’s not real time. Sometimes it’s not appropriate, you know, because when they ask a question on cough, I use the opportunity to just talk generally. So that everybody picks something, everybody gains something. Yeah, so I think that’s the next step for me. We have to go away from this platform, which seems so basic to me, and see how we can introduce AI into it and take it to the next level. So let me talk before our session, I was hearing Metaverse, you know, we have to collaborate and take from what everybody has learned. We don’t have to reinvent the wheel. Technology has come far. You know, we’re talking AI, we’re talking conversational AI. We need to collaborate and take this platform to the next level because patient outcome is important. Quality of care is important. Patient safety is important. And these are all issues that conversational AI will have an impact on. So this time next year, I don’t want to be talking about WhatsApp. I want to go for to the next level. So thank you very much.

Rajendra Pratap Gupta:
Thanks, Dr. Olavisi. And I assure you that one of the things that we promise as a tangible outcome of this year’s panel on conversational AI would be to make sure that you next time present how we help you reach the next level. That’s a very big challenge. But if you’re not able to make a difference on the ground, we are a fancy organization and we are not that. We actually mean results and we will do. So that’s why it was the reason to have you given the work that you’re doing in actual LMICs as we would call them. And if you are able to make a difference to your working as a clinician, we would have succeeded in delivering or walking our talk. Otherwise, it’s just a mere discussion, which we will not intend to. Coming to Shauna, Shauna has led a global project and we are very impressed that I know a few years back, the only project at scale for AI was IBM Watson. So Shauna, you have an experience, you have reflections. Given the journey of IBM Watson, given what we are talking now, what would be your guidance for this group for what we are talking about tangible outcomes for next year?

Shawnna Hoffman :
You know, I think that the, kind of my reflections, honestly, are this is an extremely complex problem in general. And it doesn’t have to do with just conversational AI. So as you stand back, look at all of the different aspects that makes the individual vulnerable. I think one of the concerns that I have is something that you had brought up, the 2.6 billion people who don’t have access to the internet, we need to continue to move forward with conversational AI, but we also need to make sure that those 2.6 billion people get access to the internet and to that reliable connectivity to the information. Because if we create all these chatbots and do all this amazing work and they can’t access it, then it’s really not gonna do us much good to really make that big of a difference that I know that you all want to make. I think that that would probably be my main thing that would concern me, that I would probably add beyond what the other speakers have mentioned. Because that complexity really does take it to a really tough level. And we need to look holistically at the individual and what their needs are so that they can get access and what we can do uniquely. One of the things we did with IBM Watson is to set it up in various villages to where everyone would come to one location. So there are opportunities that individuals don’t need just even a cell phone, but providing access where it’s walkable to them within a few miles or even many miles, but at least within like half a day to be able to get access to this remote medical information.

Rajendra Pratap Gupta:
Thank you, Shana. Now we will move to the questions that I see on the chat. What is the potential for training and learning best practices? So on the training side, at least what I can mention about is that we have courses on digital health at every level for doctors. I mean, there’s a postgraduate certificate course. You can look up digitalacademy.health. We have courses for health professionals, but what we are also coming up, which is very interesting is courses on AI and robotics for class eight students with IIT Delhi we have tied up, is that we need to educate people at the bottom, right from class eight onwards where they start learning about it. And this is the elementary course. And then what also we are launching early next year is the frontline health workers course. If they’re not educated, we’re not going anywhere. So that’s on the training side. On the best practices side, I would put this question to Sameer Pujari, given that WHO is probably one of the best platforms to look at, or even the Dynamic Coalition on Digital Health to look at collaborating with Sameer on the best practices on AI for conversational AI. Sameer, over to you. Can you unmute Sameer Pujari, please?

Sameer Pujari:
Yep. Hi, sorry, there was a lapse in the network connection for some reason. But I just wanna mention that on the training part, WHO has converted the guidance that they have created in the last year. And there’s an open WHO course available on ethics and governance of AI. And this is not just a theoretical course. It has a very practical checklist of an approach. And I’ve put that in the chat, the link to the course, which has been taken by more than 17,000 people across 170 countries virtually. So I think that’s one of the solutions, but I think one of the products which is there, we are coming up with a course specifically for developers, because it’s important for this community to understand what it means to create an ethical approach. And this course will be going live by the end of the year. We are having similar courses coming up with the regulations side as well. And these are targeted to developers, to policymakers, and to implementers. So there are checklists and application sort of processes for each of them within this course materials. These are being used by academic institutions across the globe to train students on healthcare provisions and AI. So these are some of the ways that it is there. But again, I keep reiterating the things. It’s not, let’s not recreate the wheel. Let’s join hands, there’s content available, and we can deploy as many ways as we can through the process. Back to you.

Rajendra Pratap Gupta:
Thanks, Sameer. Ashish, over to you. This is an interesting question. Is the role of conversational AI in dispelling superstitions and health fallacies?

Ashish Atreja:
That’s a great one. I think there is a clause that if you’re not intentional about something, then that’s not gonna happen. So which means we do know there is a lot of misperceptions in healthcare. We saw in COVID what happened. And if we just leave at this, like in social media, WhatsApp or others, there’s a lot of chance of things going viral, which are not accurate. And what we realized in COVID was clinicians, researchers actually did not have much voice because most viral content was the one which was the least trusted content from clinicians. So I think part of this is coming back to this stuff is the onus is on us to put science as a base, right? So when technology solutions are created, and because it’s democratizing technology, anyone can, within a week, learn using these technologies to create a bot and to do it. That may not be validated, and many times are not if it comes out so early. So we have to put kind of some framework. One can call it guardrails. If it is very life-threatening things, we have to put very rigorous guardrails. So FDA, Food and Drug Administration in US has a three-point system. It’s a life-threatening system thing that has to go through much more clinical evidence, multiple clinical trials. If it is moderate risk, then certain kind of a thing. If it is very low risk, then it can go without major clinical trial. So we used to have some kind of a framework like that. If it is an education content, can we even use generative AI to validate some of the content which may come out? If we create a generative AI, not on large language models on the internet, because then it will hallucinate, but can we create the large language model on Harrison’s Medical Textbook, which I got trained on? Can we train, get on WHO practices, on VA practices, open domain content from US, UK, developing countries, WHO, wherever it is, on textbooks? Then we actually may have an automated way or semi-automated way to check the accuracy of it, put some delimiters, maybe backing with human in the loop for critical things. So I think that framework is not here right now, but we need to go beyond, Dr. Gupta, as you mentioned, from traditional ways of regulating to actually maybe semi-automated bot ways of regulating. I was in a security summit and gave a keynote there and where it ended was, they’re gonna be more and more bots on the trying to hack information now. So right now, humans do this bots to kind of get into security and hacking. With generative AI, it’s gonna be bots that are gonna be doing it. So we need to dwell bots, which are gonna be protecting us in that. And so the similar thing we have to do, we may not be able to do this governance just by humans alone. We have to go one-to-many and automated governance backed by human’s loop to allow that.

Rajendra Pratap Gupta:
Thanks, Ashish. And I think the point that you raise is very important. The Dynamic Coalition for Digital Health at the UN’s IGF. One of the things I would add to what Mavish proposed was not only the generative health AI, but also generative health AI governance framework. I think if I’m sure there are multiple, but we need to come out with something which is understandable, implementable over the next year or so. I have interesting question that I would post to Sabin is, how can conversational AI technology be made more accessible to people in low-income areas who may have limited access to smartphone or the internet? I know that you did had a passing reference to this Sabin, but you would like to add something on this?

Sabin Dima:
Yeah, at least we need an internet connection if we want to have access to powerful models, but there are models very efficient that you can run it, not on a tablet, but only on a mobile phone. But I see something like the digital doctor of the village that encapsulate the knowledge from all of the doctors from all around the world. And basically you need one mobile phone for every village. So this is the minimum resource that everybody needs. I like another question, to what extent can conversational AI pose a threat to employment? I’m always saying, and I said it before, I think that AI is not going to take your job, but the human using AI will take your job for sure. Probably using AI, employers will work only two or three days per week, and we will achieve 10 times more results. In the same time, you know, that it’s a big problem in healthcare in general, that the human error and AI can supervise this. So imagine that you will do your job having maybe 100 AI assistant helping you perform better. So I don’t see any threat for employment.

Rajendra Pratap Gupta:
And I mean, I will add to that, that in the other dynamic coalition on internet and jobs, we had a session yesterday on Project Create. Create is collaborate to realize the employment and entrepreneurship for all through technology ecosystem. In fact, we have created job maps for nine sectors. And we have talked about the conventional models of doing a business and the Create model. So let’s not fear technology. I think technology is best used for creating jobs than taking away jobs. And this Project Create is about that. So I would say look up this website called projectcreate.tech. We are releasing our framework tomorrow afternoon at IGF on Project Create. So while the threat is not for jobs, the threat is to lack of competence. I would put it this way. So I would say upskill yourself, be competent. If you’re not competent, anyone can threaten you, not only AI. So I would say that please upskill, continuously upskill, crosskill yourself. And that’s important. So there is no threat to you if you are updated and upskilled. Well, if you are not, you certainly have. Sabin, you’re trying to say something?

Sabin Dima:
I agree, I agree.

Rajendra Pratap Gupta:
Thank you. Let’s look at the other questions that are there. Where can we access the recording of this conference? Is there on YouTube, IGF broadcast that on YouTube? So it’s available for people to watch. There is a comment, I guess. I believe youth-mediated initiative would help bridge the digital literacy gaps. Yes, of course. And we have, I think yesterday, we were surprised to have a digital health session by the youth tech envoy of the ITU. And she is keen to work with the DC Digital Health to address this big issue of youth’s involvement in digital health and DC. There’s another question. Ashish’s comment on I am hoping science, which is evidence-based, validated, repeatable, applicable outcomes and transparency ethical approach can help build trust along with great patient experience. Yes, Ashish, totally agree with you. And that is what I think this group should be working on, on the governance and the outcome. So by the way, on the other side with the International Standards Association, we are working on the outcomes measures using technology. I think Dr. Enkasing from my team is going to make a presentation to the meeting in Arlington, I guess it’s next month on the how to measure clinical outcomes of technology driven initiatives. And we are especially talking of digital therapeutics, which is being led by health parliament at the, it’s called the Bureau of Indian Standards, which represents the ISI, the International Standards body. So this was a great session. We are up our time. And I thank each one of you for taking our time and different time zones and enriching us on conversational AI, giving us a pathway for next year. I also thank our technical team at IGF for making this session seamless for us. Thank you all so much. And we will connect back in the mainframe and hopefully next year we’ll come back with the tangible outcomes we discussed. The goal would be Dr. Olavesi should benefit of all we talked. That would be the goal for us. Thank you so much. Thank you. Thank you very much. Thank you. Thank you. Thank you all. Thank you.