Applying Collective Intelligence to Advance Deep Learning Models

Collective intelligence can be used to improve deep learning models by incorporating principles of self-organization and adaptability.

• Deep learning networks require sophisticated engineering and careful training schemes.
• Collective intelligence produces systems that are robust, adaptable, and have less rigid assumptions about their environment configurations.
• Active areas in modern deep learning research that incorporate collective intelligence include deep reinforcement, multi-agent, and meta learning.
• An example of collective intelligence in action is the annual Reddit art Place experiment where users collaborate and coordinate a strategy to create a meaningful design.
• Analog neural networks developed in the 1980s were much closer to natural adaptive systems and produced amazing results such as object extraction.
• Collective intelligence can be applied to image processing, generative models, deep reinforcement learning, multi-agent learning, and meta learning.

An example of collective intelligence in action is the annual Reddit art Place experiment where users collaborate and coordinate a strategy to create a meaningful design.

Collective intelligence studies the group brainpower that emerges from the interactions of many individuals. It's commonly observed in nature—for example, when a group of fish decides which direction to swim or when elephants choose where to migrate. Google Brain's David Ha, a research scientist, shares methods for using collective intelligence to improve today’s deep learning models. The current generation of neural network models achieves state-of-the-art performance on tasks across fields spanning computer vision, natural language processing, and reinforcement learning. But as these models become larger and more complex, they suffer from issues including poor robustness, the inability to adapt to novel task settings, and other problems. Collective behavior, however, tends to produce systems that are robust, adaptable, and have less rigid assumptions about their environment configurations. In this keynote, Ha highlights several active areas in modern deep learning research that incorporate the principles of collective intelligence to advance current capabilities, including lessons in deep reinforcement, multi-agent, and meta learning. He will provide examples from Reddit, Conway's Game of Life, Minecraft, Puzzle Pong, and more. Ha works in the Google Brain team in Japan, where his research interests include complex systems, self-organization, and creative applications of machine learning. Prior to joining Google, Ha worked at Goldman Sachs as a Managing Director.