Generative artificial intelligence (AI) has the potential to revolutionize healthcare by advancing drug development and enabling quicker diagnoses. However, the World Health Organization (WHO) has raised concerns about the rush to embrace AI, warning of potential drawbacks. The WHO is assessing the potential dangers and benefits of large multi-modal models (LMMs) powered by AI, which are becoming increasingly popular in the healthcare sector.
Generative AI uses algorithms trained on data sets to create new content, while LMMs can accept multiple forms of input, such as text, images, and video, to produce diverse outputs. According to Alain Labrique, WHO’s digital health and innovation director, some experts believe LMMs can imitate human thinking, behavior, and problem-solving methods. They are expected to have broad applications in medicine, public health, scientific research, and drug development.
The WHO has identified five key areas where LMMs could be utilized: diagnosis, scientific research, medical and nursing education, clerical tasks, and patient-guided use. Despite their potential, the WHO cautioned that LMMs might yield false, biased, or incomplete results due to poor quality data or biases related to factors like race, ethnicity, gender identity, or age.
The organization warned of potential “automation bias,” where users might indiscriminately rely on the algorithm despite legitimate reasons to doubt it. To address these issues, the WHO issued recommendations on the ethical and governance aspects of LMMs, emphasizing the need for transparent information and policies to manage their development and application.
While AI has been utilized in public health and clinical medicine for over a decade, the booming use of LMMs poses new risks, according to the WHO. Concerns were raised about existing regulations, data privacy, and the influence of tech giants in developing LMMs. The WHO recommended that LMMs be developed with input from medical professionals and patients, ensuring data privacy and providing patients with an option to opt-out.
Additionally, the WHO highlighted cybersecurity threats to patient information and healthcare provision that could arise from LMMs. The guidance provided by the WHO aims to ensure that AI can contribute to humanity’s well-being while adhering to the highest ethical standards. Jeremy Farrar, WHO’s chief scientist, stressed the importance of identifying and addressing the associated risks to maximize the potential benefits of generative AI in healthcare.