A Journey to the Future of AI

A Journey to the Future of AI, by Prof. Pieter Abbeel

FRI 06: 10.45 – 11.30

AI speaker Pieter Abbeel is an AI & Robotics Professor, as well as the Robot Learning Lab’s Director at UC Berkeley. In 2014, Pieter co-founded Gradescope.com, while in 2017, he co-founded Covariant.ai. He is also an Advisor for many companies in the Robotics and AI field, including OpenAI. Moreover, Pieter created the Venture Fund AI@TheHouse and often gives lectures on artificial intelligence.

Pieter started the Berkeley Robot Learning Lab following his appointment at UC Berkeley as an assistant professor. Furthermore, in 2014, he helped develop Gradescope, which was sold to TurnItIn in 2018. Pieter created Gradescope with Sergey Karayev, Arjun Singh, Ibrahim Awwal, and several other UC Berkeley engineers.

Speaker Pieter Abbeel began working for OpenAI in 2016. Since then, he has written plenty of articles on robot learning, reinforcement learning, and unsupervised learning. 2016 saw him also take on the role of co-director of the Berkeley Artificial Intelligence Research Lab. The BAIR involves undergraduate and postdoctoral students that are passionate about robotics and machine learning.

His other venture is Berkeley Open Arms, which has licensed the intellectual property (IP) from Berkeley related to the Blue Robot project. In 2017, he was appointed as a tenured, full-time professor at the University of California, Berkeley.  Pieter Abbeel, Rocky Duan, Peter Chen, and Tianhao Zhang launched AI-based company covariant.ai in 2017. The New York Times, MIT Technology Review, Wired, and IEEE Spectrum all wrote articles regarding the company’s launch.

Covariant.ai leverages recent developments in deep reinforcement learning and deep imitation learning to produce artificial intelligence software that simplifies the process of teaching robots new, advanced skills. Abbeel is actively performing researches as well as teaching upper-division and graduate courses in robotics, artificial intelligence, and deep unsupervised learning.

Pieter’s work aims to get robots to learn from humans and from trial and error. He also works on speeding up skill acquisition using meta-learning. Aside from sophisticated helicopter aerobatics, basic assembly, knot tying, and laundry organization skills, Pieter’s robots have learned locomotion and vision-based robotic manipulation techniques.

He received several awards, one being the best paper awards at ICML, ICLR, ICRA, and NIPS. Pieter also obtained awards from Darpa, NSF, AFOSR, ONR, TR35, Sloan, IEEE, and the Presidential Early Career Award for Scientists and Engineers. The BBC, New York Times, Bloomberg, Wired, Wall Street Journal, Tech Review, Forbes, and NPR often mention Abbeel’s work.

Speaker

Pieter Abbeel

Professor in Artificial Intelligence, Robotics at UC Berkeley