GENEVA: Generative artificial intelligence could transform health care through things like drug development and more rapid diagnoses, but the World Health Organization stressed Thursday more attention should be paid to the risks.
The WHO has been examining the potential dangers and benefits posed by AI large multi-modal models (LMMs), which are relatively new and are quickly being adopted in health.
LMMs are a type of generative AI which can use multiple types of data input, including text, images and video, and generate outputs that are not limited to the type of data fed into the algorithm.
“It has been predicted that LMMs will have wide use and application in health care, scientific research, public health and drug development,” said the WHO.
The United Nations’ health agency outlined five broad areas where the technology could be applied.
These are: diagnosis, such as responding to patients’ written queries; scientific research and drug development; medical and nursing education; clerical tasks; and patient-guided use, such as investigating symptoms.
While this holds potential, WHO warned there were documented risks that LMMs could produce false, inaccurate, biased or incomplete outcomes.
They might also be trained on poor quality data, or data containing biases relating to race, ethnicity, ancestry, sex, gender identity or age.
“As LMMs gain broader use in health care and medicine, errors, misuse and ultimately harm to individuals are inevitable,” the WHO cautioned.
On Thursday it issued recommendations on the ethics and governance of LMMs, to help governments, tech firms and health care providers safely take advantage of the technology.
“Generative AI technologies have the potential to improve health care but only if those who develop, regulate and use these technologies identify and fully account for the associated risks,” said WHO chief scientist Jeremy Farrar.
“We need transparent information and policies to manage the design, development and use of LMMs.”
The WHO said liability rules were needed to “ensure that users harmed by an LMM are adequately compensated or have other forms of redress.”
AI is already used in diagnosis and clinical care, for example to help in radiology and medical imaging.
WHO stressed however that LMM formats presented “risks that societies, health systems and end-users may not yet be prepared to address fully.”
This included concerns as to whether LMMs complied with existing regulation, including on data protection — and the fact they were often developed by tech giants, due to the significant resources required, and so could entrench these companies’ dominance.
The guidance recommended that LMMs should be developed not just by scientists and engineers alone but with medical professionals and patients included.
The WHO also warned that LMMs were vulnerable to cyber-security risks that could endanger patient information, or even the trustworthiness of health care provision.
It said governments should assign a regulator to approve LMM use in health care, and there should be auditing and impact assessments.