Start your day with intelligence. Get The OODA Daily Pulse.

Is your head still spinning from all the AI news of late? It will all seem so quaint when the useful humanoid robots arrive.

This post tracks some of the recent breakthroughs in humanoids, with a focus on the humanoid robots with Embodied AI.

Let me start with a prediction: By the end of 2024, humanoid robots with Embodied AI will be able to perform useful tasks at scale. They will be proven as useful in manufacturing, warehousing, store operations/restocking and hospital/healthcare operations. By the end of 2025, humanoid robots will proliferate into homes.

We have all been watching the rise of automation, but few will be prepared for this exponential shift.

Historical Context on Embodied AI and Humanoid Robots

Embodied AI was first used to describe the special branch of AI that focuses on how systems interact with the physical world, including AI for sensorimotor skills, autonomous navigation, real-world interactions and learning. Recently it is taking on a broader meaning. With the rise of Generative AI and the multiple research paths building towards multi-function AI, the term Embodied AI is being used to capture the concept of a robot that gives physical form to AI. A robot is the body for AI.

I have tracked humanoid robotics closely since my years as an enterprise technologist at TRW, Northrop Grumman and in government. Just over a decade ago the most advanced humanoid robotics projects were either activities funded by NASA (the most famous of which was their Robonaut project) or demonstration projects like the Asimo by Honda. Neither did anything really useful. To date, all the interesting humanoid robot demos seen on YouTube or TV have yet to do anything useful either. But the pace of news and increasing capabilities are pointing to significant progress.

Several big tech trends have enabled these advancements. Batteries are more powerful and last longer, new types of actuators require less power and weigh less, compute is more powerful, sensors are better, and new types of AI (especially deep learning) are available (many of these trends were examined in detail at OODAcon).

Developments Over The Last Six Months

In just the last six months humanoid robots have been fielded as prototypes into environments where they are co-working with humans. Humanoid robots in labs have been demonstrated with abilities to handle complex tasks like cooking, cleaning folding clothes, soldering, and making coffee. The learning by humanoid robots with embodied AI is now accelerating to the point where robots can learn by watching what humans do.

Examples of note:

  • 1X, a company backed by OpenAI, as revealed a humanoid robot based on a single tech stack (a single neural network) that works autonomously to accomplish multiple dynamic tasks never before accomplished by humanoid robots.
  • Boston Dynamics’ Atlas was shown performing increasingly complex gymnastics and even artful parkour over complex obstacles.
  • Amazon began real world trials of Agility Robotics new Digit humanoids, which are designed to work around people.
  • Tesla’s Optimus continued to improve with new demonstrations of increased agility, increased ability to navigate around people in dynamic workplaces and an ability to handle delicate tasks like handling eggs.
  • Robot builder Sanctuary.ai demonstrated robots performing increasingly complex and delicate tasks, like soldering, welding, cooking, cleaning, restocking shelves.
  • Clone robotics unveiled a soft robotic hand with amazing dexterity and sensing capabilities.
  • Singapore based Forex Intelligence released a new production version of their GR-1 robot. They plan on producing these in mass.
  • Limx Dynamics continued production of both humanoid and other advanced mobile robotics. Their general purpose CL1 has begun testing and is expected to enter service this year. It has its own vision processing capability.
  • Figure, a company that did not exist 18 months ago, demonstrated how Embodied AI with an advanced end-to-end neural network can learn to do things just by watching humans. (The demo video below captures this very amazing feat)

Concluding Comments/Recommendations:

If you are an enterprise CIO, CTO, CISO, you should realize you will have a role in operating and maintaining all your organizational robots, including those humanoid robots with embodied AI.

We expect the most rapid uptake of the use of humanoid robots with embodied AI will be in the manufacturing sector. They will also be of increasing use in warehouses. Other use cases will come after that. If you are in these sectors see what you can do to accelerate the use of humanoids in your operation so you can use them to business advantage (if you don’t you will be disadvantaged because others will). If sell to these markets understand how this shift will change your customer buying patterns.

We mentioned the 9 companies because of what we have seen them producing, they are exemplars we believe our community should track. But review of Crunchbase shows over 12,000 companies involved in this sector. This is a very dynamic market with many firms producing in stealth. All of us should be ready for market changing surprises in this fast moving field.

Bob Gourley

About the Author

Bob Gourley

Bob Gourley is an experienced Chief Technology Officer (CTO), Board Qualified Technical Executive (QTE), author and entrepreneur with extensive past performance in enterprise IT, corporate cybersecurity and data analytics. CTO of OODA LLC, a unique team of international experts which provide board advisory and cybersecurity consulting services. OODA publishes OODALoop.com. Bob has been an advisor to dozens of successful high tech startups and has conducted enterprise cybersecurity assessments for businesses in multiple sectors of the economy. He was a career Naval Intelligence Officer and is the former CTO of the Defense Intelligence Agency.