Page 164 - Kaleidoscope Academic Conference Proceedings 2024

2024 ITU Kaleidoscope Academic Conference

frameworks and standards to support the trustworthy empowers individuals everywhere to live healthier lives. As
deployment of AI technologies like ours in clinical we stand at the cusp of a new era in healthcare innovation,
environments. Despite these challenges, our work we have a tremendous opportunity and responsibility to
demonstrates the vast potential of generative AI and harness the transformative potential of AI to benefit
knowledge retrieval to transform health services and humanity as a whole. This research serves as a step towards
experiences. By enabling more personalized, accessible, and that vision, laying the foundation for personalized, scalable,
engaging interactions, these technologies can empower and trusted AI health services. We hope it spurs further work
individuals to take greater agency over their health and to meaningfully bridge the gap between cutting-edge AI and
wellbeing. As we continue advancing the state-of-the-art in real-world health impact.
natural language AI and its application to healthcare, it is
imperative that we do so responsibly, with a clear focus on REFERENCES
benefiting patients and augmenting human care capabilities.
With the right technical and institutional safeguards in place, [1] Arshi, T. A., Ambrin, A., Rao, V., Morande, S., &
we believe AI-powered health systems can make quality, Gul, K. (2022). A Machine Learning Assisted
proactive, and preventive care more available to all. Study Exploring Hormonal Influences on
Entrepreneurial Opportunity Behaviour. Journal of
6. CONCLUSION Entrepreneurship, 31(3), 575–602.
https://doi.org/10.1177/09713557221136273
The rapid advancements in artificial intelligence, particularly
generative AI and large language models, have unlocked [2] United Nations. (2015). Sustainable development
unprecedented opportunities to transform the delivery of goals. SDGs Transform our world, 2030.
personalized digital health services. As highlighted by the
United Nations' Sustainable Development Goals and ITU's [3] ITU. (2021). Harnessing the power of technology
vision, harnessing AI-driven technologies can significantly in Least Developed Countries.
contribute to bridging healthcare disparities and fostering https://www.itu.int/hub/2021/04/harnessing-the-
human wellbeing worldwide. This research presented a power-of-technology-in-least-developed-countries/
novel framework that leverages the synergies between
generative AI, knowledge retrieval, and domain expertise to [4] Brown, T., Mann, B., Ryder, N., Subbiah, M.,
enable intelligent, user-centric health support at scale. Our Kaplan, J. D., Dhariwal, P., Neelakantan, A.,
proposed architecture seamlessly integrates state-of-the-art Shyam, P., Sastry, G., Askell, A., Agarwal, S.,
language models, curated biomedical knowledge bases, and Herbert-Voss, A., Krueger, G., Henighan, T.,
intuitive user interfaces to provide individuals with Child, R., Ramesh, A., Ziegler, D., Wu, J., Winter,
personalized, trustworthy, and engaging health interactions. C., ... Amodei, D. (2020). Language models are
By dynamically augmenting generative models with relevant few-shot learners. Advances in Neural Information
domain knowledge, our approach aims to ensure AI-driven Processing Systems, 33, 1877–1901.
health recommendations are not only fluent and contextual,
but also clinically valid and grounded in scientific evidence. [5] Morande, S., Del Vacchio, E., & Ranieri, A.
(2020). Digital innovations in healthcare startups:
The extensive empirical evaluations, spanning automated transforming service ecosystem. Journal of
benchmarks and user studies, demonstrate the promise of this Business Management Studies, 2(1), 26–39.
direction. We show that retrieval-augmented generation can
help produce health information that is highly relevant to [6] Preiksaitis, C., & Rose, C. (2023). Opportunities,
users' specific needs and circumstances, while maintaining challenges, and future directions of generative
strong alignment with established medical facts and artificial intelligence in medical education:
guidelines. The positive feedback from patients and scoping review. JMIR Medical Education, 9,
providers alike underscores the practical utility of our e48785. https://doi.org/10.2196/48785
approach in real-world settings. However, we also highlight
the substantial challenges and open questions that remain in [7] Morande, S., & Pietronudo, M. C. (2020).
responsibly building and deploying generative AI for Pervasive Health Systems: Convergence through
healthcare. Our discussion emphasizes key considerations Artificial Intelligence and Blockchain
around safety, robustness, fairness, privacy, and Technologies. Journal of Commerce and
interpretability - all of which are especially critical given the Management Thought, 11(2), 155.
high-stakes nature of health applications. We advocate for https://doi.org/10.5958/0976-478x.2020.00010.5
continued research and multidisciplinary collaboration to
tackle these issues. Ultimately, the success of AI-powered [8] Topol, E. J. (2019). High-performance medicine:
health systems will depend not only on technical advances, the convergence of human and artificial
but also on fostering public trust and value alignment. By intelligence. Nature Medicine, 25(1), 44–56.
prioritizing patient agency, clinical validity, and ethical https://doi.org/10.1038/s41591-018-0300-7
development practices, we can work towards a future where
AI equitably extends the capacity of health systems and

– 120 –