Value Compass

Aligning AI with Basic Human Values

To push the frontier of AI towards a human-machine symbiotic future, we propose the interdisciplinary value compass project, aiming to achieve LLM alignment based on the basic values from humanity and social science. Three parts are mainly explored:1) efficient and adaptive alignment algorithms; 2) reliable value evaluation systems; and 3) interdisciplinary studies on values and machine ethics. Hoping to create a new generation of AI systems that compatible with human values of various cultural contexts.

Contact Usvaluecompass@microsoft.com Git Hub