WHO Calls for Caution on AI in Healthcare to Avoid Widening Global Divide
• WHO issues new guidelines on use of large multi-modal AI models in healthcare, warning of potential dangers for poorer countries if models only trained on data from wealthy nations
• Rapid growth of generative AI capabilities prompts WHO to update existing AI guidelines less than 3 years after first set introduced
• Possible risks highlighted include propagating inequities, model collapse from training on inaccurate data, industrial capture of development, and race to release unsafe applications
• Recommendations include mandatory audits, ethics training for developers, and involvement of civil society groups and healthcare recipients in oversight and regulation
• If not cooperatively regulated by governments, powerful generative health AI could be shaped mainly by interests of tech companies and wealthy countries