Conference:  Transform X 2022

Authors: Pieter Abbeel

2022-10-19

Pieter Abbeel wears many hats: Professor at UC Berkeley, Director of the Berkeley Robot Learning Lab, Founder of three companies, podcast host, and investor. The common thread is that Professor Abbeel is passionate about AI and robotics. In this keynote presentation, he explores the possibility of training a large neural network to enable faster learning in robotics. Professor Abbeel discusses his lab’s approach to solving this problem and will cover how video prediction is an excellent proxy for generalizable robots, the relevant models and datasets useful for pre-training, how unsupervised learning can help robots learn from themselves; and the usefulness of a human-in-the-loop. He describes a four-step framework that might be able to lead, ultimately, to generalized robotics. Professor Abbeel is co-director of the Berkeley Artificial Intelligence (BAIR) Lab and founded Gradescope, which provides AI to help instructors with grading homework and exams, and Covariant, which provides AI for robotic automation of warehouses and factories. He is also a founding partner at AIX Ventures, a venture capital firm focused on AI start-ups, and is the host of The Robot Brains podcast, which explores what AI and robotics can do today and where they are headed.

Conference:  Transform X 2022

Authors: David Ha

2022-10-19

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.