TechCrunch’s robotics newsletter Actuator has exciting plans for the following weeks, featuring insightful Q&As with some of the top minds in the field. The roster of experts includes renowned individuals such as Matthew Johnson-Roberson from CMU, Max Bajracharya and Russ Tedrake from Toyota Research Institute, Dhruv Batra from Meta, and Aaron Saunders from Boston Dynamics.
One of the featured experts, Ken Goldberg, a professor at UC Berkeley and co-founder of Ambidextrous, is set to share his valuable insights on the role of generative AI in the field of robotics. Goldberg firmly believes that generative AI, specifically large language models like ChatGPT, will revolutionize communication between humans and robots. He also mentions how Vision-Language-Action models are currently being developed to enhance robot perception and control of robot arms and legs.
Goldberg emphasizes the importance of collaboration among robotics labs to share data, as it has been driving progress in generative AI. However, there are still questions surrounding the generalization of these models.
In his discussions, Goldberg highlights the significance of “Multi-Modal models,” which involve combining different input modes such as vision, language, tactile abilities, depth sensing, and robot actions. He also discusses the concept of Diffusion Policies, developed by Shuran Song, as an exciting way to preserve multi-modal actions.
Touching upon recent advancements in the robotics industry, Goldberg mentions companies like Boston Dynamics, Agility, and Unitree, which have made significant strides in the development of humanoid robots. He acknowledges the advantages of legged robots over wheeled ones when it comes to traversing diverse environments.
While Goldberg recognizes the importance of bimanual robots, he believes that simple grippers will continue to be more reliable and cost-effective. He also highlights the potential increase in robot utilization in manufacturing and warehouses following recent union wage settlements.
The progress of self-driving taxis is undeniable, but questions about their cost-effectiveness still linger. Goldberg also delves into the realm of robot-assisted surgery and explores the concept of “Augmented Dexterity” to enhance surgical skills.
Looking ahead, Goldberg does not anticipate the emergence of true AGI (Artificial General Intelligence) or general-purpose robots in the near future. However, he predicts that within the next decade, affordable home robots will be capable of decluttering homes by picking up clothes, toys, and trash.
Despite the vast scope of robotics, Goldberg believes that the importance of robot motion planning is often overlooked. He highlights the issue of robot singularities, where robots unexpectedly stop and require manual resetting, as a fundamental problem in motion planning.
To address this problem, Goldberg co-founded Jacobi Robotics, a startup that focuses on developing efficient algorithms to avoid singularities. Their efforts aim to increase reliability and productivity for all types of robots.
With such knowledgeable experts and groundbreaking advancements in the field of robotics, Actuator’s Q&As promise to be an enlightening and informative series for all robotics enthusiasts.