Cheriton School of Computer Science Professor Kate Larson and her international colleagues have published a commentary in Nature about the need for cooperative artificial intelligence — beneficial AI with social understanding.
This need is particularly urgent as in the coming years AI systems that interact quickly and complexly with each other and with humans will become ever more common. From city streets and roads to consumer and financial markets, from mass communications technologies to cyber and physical security systems, AI systems that engage poorly with humans not only will fail to deliver their promised benefits, but they may also disrupt personal and societal relationships in disastrous ways.
What’s needed is a science of cooperative AI — one that prioritizes development of cooperative intelligence with the ability to promote mutually beneficial joint action.
Professor Larson and her colleagues believe that cooperative intelligence is unlikely to emerge as a by-product of research on other kinds of AI. What’s needed is more research on cooperative games and complex social spaces, on understanding norms and behaviours, and on social tools and infrastructure that promote cooperation.
To succeed, cooperative AI must connect with the broader science of cooperation, a field that spans the social, behavioural and natural sciences. AI researchers will need to converse and work with researchers across a variety of fields — psychology, law and policy, history, sociology and anthropology, and political science and economics — to develop AI that aligns fairly with human intentions, preferences and values.
Read their full commentary in Nature (PDF).
Citation
Allan
Dafoe,
Yoram
Bachrach,
Gillian
Hadfield,
Eric
Horvitz,
Kate
Larson,
Thore
Graepel.
Cooperative
AI:
machines
must
learn
to
find
common
ground.
Nature,
593,
33–36
(2021).