Notes on robotics research

Actuator: UC Berkeley/Ambi Robotics' Ken Goldberg discusses the biggest robotics trends of 2022

Happy holidays from the Ghost of Actuator Past. I’m writing you from the beginning of the month and hopefully not checking my work email or Slack as you’re reading this. I’ll be back in action next week (hopefully the bags under my eyes will have subsided slightly), but until then, I’ve got one more great interview for you. This week I leave you in the very accomplished hands of Ken Goldberg, who dons the dual hats of U.C. Berkeley robotics professor and chief scientist at Ambi Robotics.

Q&A with Ken Goldberg

Ken Goldberg at TechCrunch Disrupt

Image Credits: Kimberly White (opens in a new window) / Getty Images

TC: What was the biggest robotics story of 2022?

KG: For me, three major robotics developments in 2022 stand out:

  1. The surprising progress of large language models (e.g., GPT-3) and associated text-to-image generation (e.g., Dall-E) is spurring excitement in the robotics community about how these can be applied to robotics, by completing robot-relevant prompts. An exciting paper from Brian Ichter and colleagues at Google AI was presented at the 2022 Conference on Robot Learning on December 14–18.
  2. Elon Musk reframing Tesla as a robotics company, with their associated research initiative into humanoid robots. I doubt they will build a useful humanoid robot for $20K in 2 years, but Tesla has great expertise in sensors/motors, robot use cases in its own factories, and awareness of costs and mass production that is a strong vote of confidence in the field.
  3. The widespread adoption of robots in warehouses to meet increasing demand for e-commerce was not diminished by the return to stores or inflation. Companies such as Ambi Robotics have installed 70+ AI-powered robotic sorting systems across the U.S., demonstrating the viability of deep learning to enhance worker productivity.

What are your biggest robotics predictions for 2023?

I believe we’ll see robots-as-a-service (RaaS) models that make robots available to a much wider segment of industry (e.g., Model T Ford financing that opened up car-buying to the middle class).

How profound of an impact has the pandemic had on robotics?

The pandemic dramatically increased adoption of teleconferencing and also telerobots in hospitals and nursing homes to prevent virus transmission. But the biggest impact was on e-commerce, which grew at 5x the previous pace.

How much of an impact has the macroeconomic environment had on robotics investing?

Venture capital is far harder to obtain than it was in December 2021, but funds are continuing to make investments in robotics companies with growing demand like Locus and Ambi.

What underaddressed category deserves more focus from robotics startups and investors?

Robotics-as-a-service (RaaS) shifts the cost from capital expenditure to operating expenditure, so it is very attractive and practical for industry.

How will automation impact the workforce of the future?

As Diego Kuonen noted: “It’s not about replacing the human with a robot. It’s about taking the robot out of the human.” Robots cannot replace most of the dextrous and nonrepetitive aspects of work. Robots won’t replace people; they will increase worker productivity.

Are home robotics finally having their moment?

Designing a cost-effective home robot to do more than cleaning floors is extremely challenging. Aging demographics will increase demand for this but it will take more time.

What more can/should the U.S. do to foster innovation in the category?

Fortunately, National Science Foundation budgets were increased this year; the NSF provides critical support for graduate students and increases much needed diversity.