Subject: Complementary AI Initiative

Description: Exploring how to train AI models that augment rather than mimic human capacity.
The notion of an Artificial Intelligence (AI) that computationally mimics human intelligence has dominated the design of computational intelligence, but is tuned to replace humanity rather than augment its capacity. Rather than targeting Artificial General Intelligence (AGI), we argue that AI should be designed to complement individual and collective human capacity, avoid humanity’s blind spots, overcome cognitive and cultural limitations, and unleash novel senses, skills, and imagination.

This initiative addresses new and complex questions regarding how to build AI models without targeting but rather expanding human capacity. This research initiative will integrate with social and behavioral sciences to identify and monitor changing human limitations and opportunities. It will also draw on business and economics to measure and model human-AI productivity; legal and policy expertise to propose models by which AI can assist in the regulation of other AI; and tools and insight across the sciences, engineering, arts, and humanities to align AI with human needs and values. These efforts will grow human-AI potential to grapple with the greatest challenges facing humanity together.