Are ants intelligent? Watching an individual ant carry a bit of leaf back to the anthill, it may not seem that way, but as a group, the colony exhibits a kind of collective intelligence, says SFI’s Melanie Mitchell.

“If you look at the ant colony, each individual ant is not very intelligent and can’t do much on its own, but working together, the hundreds of thousands of different ants in a colony can do all kinds of seemingly intelligent things, like building elaborate structures underground that regulate temperature and humidity,” says Mitchell, the Davis Professor of Complexity at SFI, who conducts research on visual recognition and conceptual abstraction in AI systems.

Insects’ collective intelligence was just one of the topics of a workshop Mitchell co-organized and participated in last August. The gathering, held virtually due to the pandemic, pooled knowledge from biologists, computer scientists, and other experts to further the conversation about collective intelligence research and how it can inform AI. Some of the questions the group explored were: What mechanisms allow collective intelligence to emerge from a group of individuals or components? Which AI research paths hold the most promise for solving complex collective problems like climate change or epidemics? How can we harness collaborative intelligence to increase fairness and other key values?

Highlights of the workshop, part of SFI’s Foundations of Intelligence project, included a talk by Jeff Hawkins, co-founder of the AI organization Numenta, on how communication between columns of neurons in the brain is a form of collective intelligence, a discussion on resolving semantic ambiguities that hindered learning across fields, a debate about how to improve the study of democracy, and a discussion on the spread of misinformation.

“One of the things I took away from several of the talks was the need to tune individual agents and the conditions under which they interact in order to enhance their collective intelligence,” says SFI postdoctoral fellow Tyler Millhouse, who also helped organize the workshop along with SFI External Professor Melanie Moses. For example, biologist Anna Dornhaus noted in her talk that in insect colonies, more information-sharing between individuals isn’t necessarily beneficial to the colony as a whole. “These biologically-inspired ideas illustrate an area for improving collective AI that might not have been obvious otherwise.”

This workshop was part of the ongoing Foundations of Intelligence project, which has included Foundations of Intelligence in Natural and Artificial Systems (March 15-19, 2021), Frontiers of Evolutionary Computation (July 21-23, 2021), Can Algorithms Bend the Arc Toward Justice? (March 30-1, 2022), and Embodied, Situated, and Grounded Intelligence: Implications for AI (April 12-15, 2022).