In early 2023, OpenAI’s artificial intelligence language model, ChatGPT, stunned the medical community by performing remarkably well on all three parts of the United States Medical Licensing Exam without any specialised training. Google’s Med-PaLM 2 also achieved ‘expert-level’ performance on similar tasks, reflecting AI’s capacity to assimilate and utilise extensive medical knowledge. In a simulated setting, ChatGPT even surpassed human physicians in delivering empathetic responses to patient queries on a social media forum.
The transformative power of AI extends beyond the clinical setting. Researchers from McMaster University and MIT harnessed AI to discover new antibacterial molecules effective against the multi-drug resistant bacteria Acinetobacter baumannii. This is a significant stride in combating antimicrobial resistance, one of this century’s greatest medical challenges.
There is a rise of AI-powered chatbots aiming to answer medical queries as effortlessly as using Google today. As these algorithms become increasingly sophisticated, they could integrate data from numerous sources, thereby enhancing their accuracy. These data could span beyond language, embracing dermatological appearances, imaging, pathology, health records and genomics. Such advancements integrated into existing technologies — like the digital health platform Halodoc — could revolutionise healthcare accessibility, including in Indonesia.
Indonesia’s physician-to-population ratio is three times lower than that of East Asia and the Pacific, a problem acutely felt by over 42 per cent of the population living in rural areas. Combining Indonesia’s rich demographic and health data with AI’s analytical capabilities can aid doctors in diagnosing and treating patients more accurately. This could potentially raise healthcare services quality across the archipelago.
Consider the case of CognoSpeak, an AI tool developed by researchers from the University of Sheffield. This web-based system analyses patients’ language and speech patterns to detect early signs of dementia, promising much faster and more accurate assessments than current methods. These advances can transform the dementia care pathway in the Asia Pacific region, home to an estimated 23 million people living with dementia according to Alzheimer’s Disease International.
But certain concerns must be addressed as Indonesia seeks to leverage these technologies.
Bias stands at the forefront. AI models such as ChatGPT are primarily trained on freely-available medical literature and their outputs are only as reliable as their training data. Most of these data are sourced from studies in high-income countries and top-tier academic institutions, which may limit their generalisability.
Indonesia must consider this bias when integrating AI with existing digital healthcare services. It is crucial to generate and utilise local data — for example from platforms like Halodoc — for the effective deployment of AI in Indonesian healthcare.
Data acquisition for AI model training also invites serious privacy concerns, a high-stakes issue in Indonesia. The public’s unease over frequent sensitive data leaks — notably the 2021 National Health Insurance (BPJS Kesehatan) incident — remains high. Although the enactment of the Personal Data Protection Law in 2022 has sparked hope, there is still a further two-year time frame to formulate derivative regulations.
These developments need to acknowledge the unique challenges presented by AI in healthcare data handling, such as the pressing need for robust digital infrastructure and cybersecurity. To build its national AI-based healthcare data ecosystem, Indonesia can learn from South Korea’s model of a centralised, secure and ethical data-sharing system that assures privacy while leveraging AI’s potential. Future laws and regulations should anticipate and address potential AI applications, including health data used in AI research and AI medical chatbot operations. Establishing studies and guidelines as the groundwork for future AI healthcare regulations — such as those done by the EU and Singapore — is an excellent starting point.
The challenges of AI-enhanced healthcare necessitate a reimagining of healthcare education. Rather than fostering fear about AI replacing human roles, future professionals must be prepared to synergise with AI. This will require developing AI-focused modules in medical and health curricula, similar to China’s integration of AI education in its national strategy. Shinjini Kundu likened this transformation to the impact of automation on an airplane pilot’s role and education several decades ago.
Medical licensing exams should also evolve from emphasising right or wrong responses through multiple choice questions to assessing students on qualities that AI cannot replicate. These include a robust work ethic, respect for patients and colleagues, empathy and a deeper understanding of socioeconomic structures. The focus of medical training might shift from biology to psychology and sociology in the era of AI.
By promoting stringent regulations, prioritising data privacy and revolutionising healthcare education, Indonesia can pave the way for a future where AI and human expertise harmoniously work together. This envisioned future stands to gain from the best of both worlds — the precision and efficiency of AI coupled with the compassion and intuition intrinsic to human healthcare.
Rezka Dwi Fathana is a Master’s of Clinical Neuroscience student at University College London.