Work Hard, Play Hard: Finding Balance

The Expanding Landscape of Industrial Robotics

Traditional industrial robotics are often characterized by their substantial size, considerable weight, and potential safety hazards. Furthermore, programming these systems can be remarkably challenging, especially when adapting them to tasks beyond their original design specifications.

A significant opportunity exists for an innovative AI or software firm to streamline the programming process and enhance the versatility of these robust systems, commonly utilized in sectors like automotive manufacturing. The increasing importance of robots in modern business operations suggests that the market could potentially accommodate multiple companies specializing in this area.

Recent Developments and Investment

This week brought forth two noteworthy announcements from companies operating within this evolving space. Covariant, for instance, revealed an $80 million funding round, closely following their $40 million Series B announcement from May 2020.

An interview with Pieter Abbeel, president, chief scientist, and co-founder of Covariant, provided further insights into current robotics trends. Details of this discussion can be found here.

Increased Investment in Robotics and Automation

The surge in investment within the robotics and automation sectors since the pandemic’s onset is anticipated to not only persist but to accelerate dramatically. While demand was already present, the pandemic amplified the need for resilient and dependable robotic solutions. This acceleration was also influenced by the growth of e-commerce and Amazon’s commitment to efficient delivery services.

University Incubation and Entrepreneurship

How have universities’ approaches to incubating companies changed?

The transition from AI research to practical application has been exceptionally rapid. New ideas can be implemented into systems almost immediately after publication. This unique characteristic positions AI researchers favorably for building novel applications.

Many universities are now integrating artificial intelligence into a wider range of academic programs. For example, the “Business of AI” course at Berkeley’s Haas Business School provides students with a comprehensive understanding of AI’s current role and future potential.

To encourage greater entrepreneurship in the U.S., streamlining the visa and green card processes for international students – many of whom are leading AI researchers – would be highly beneficial.

Covariant’s Future and Addressing Bias

Will Pieter Abbeel continue teaching as Covariant expands?

Yes, a strong synergy exists between leading academic AI research at Berkeley and driving industrial R&D at Covariant. The company’s culture, fostered by CEO Peter Chen, prioritizes curiosity and lifelong learning.

How does the team address potential biases in its AI work?

Bias in AI systems is a widespread industry concern. Currently, bias doesn’t directly impact Covariant’s robotic warehousing efforts. However, quality assurance is central to all operations, encompassing data source quality, SKU performance, and customer variations – creating technical parallels to bias mitigation.

Hardware Evolution and Intrinsic’s Entry

Is hardware evolving alongside software improvements?

While Covariant primarily focuses on software and AI, they collaborate with partners to deliver complete robotic systems. Continuous improvements in hardware, particularly in end-of-arm tooling, are readily apparent. Longer-term roadmap ideas for robotic arm form factors are also under development.

Another significant development this week was the unveiling of Intrinsic, Alphabet’s latest robotics initiative. This spinout from Alphabet X has been in development for approximately five years, following a somewhat inconsistent robotics history for Alphabet/Google, including a brief ownership of Boston Dynamics. However, Intrinsic’s approach appears well-suited to Google’s core strengths.

Wendy Tan-White, CEO of Intrinsic and formerly Alphabet’s VP of Moonshots, shared her vision for the company.

Athletic Robots and Olympic Showcase

The week’s roundup concludes with a look at athletic robots. Cassie, Oregon State University’s bipedal robot, has returned to prominence after a period focused on the delivery robot Digit. Researchers successfully taught Cassie to run using a deep reinforcement learning algorithm.

Cassie completed a 5K run in 53 minutes, a remarkable achievement for a robot powered by a single battery, despite minor setbacks like an overheated computer and a navigational error.

Toyota, similarly, has been eager to showcase its technology, particularly its basketball robot CUE, at the delayed summer Olympics. CUE debuted in 2018 and returned to sink three-pointers during half-time at the USA-France game.