Future: Universal Robots or Humanoid Robots?

Last week, Google DeepMind released the RT-X model. This is a comprehensive robot model whose main goal is to expand its capabilities and apply them to different robots and tasks, including action, vision, and language understanding. This expands its practicality as a general tool in robot research and applications, in short, a universal robot.

Universal purpose robots have many uses, but we see a lot of work being done on humanoid robots. Combinations of general-purpose humanoid robots like Unitree's H1 robot are emerging, designed to perform various tasks like humans.

Another example is Tesla's Optimus, which recently underwent a major upgrade. It is one of many bipedal humanoid robots that the company is training to perform tasks they describe as "unsafe, repetitive, or boring."

On the other hand, specialized robots perform specific tasks. Most specialized robots are used for cleaning. Recently, Bangalore-based Beta Tank Robotics Pvt. Ltd. manufactured a robot for cleaning petrochemical tanks.

General vs. specialized robots

"I think 2023 will be the year of scaling up robots," said Jim Fan, who compared this paper to ImageNet, which sparked the computer vision deep learning revolution.

RT-X was trained on a massive dataset called Open X-Embodimen. This repository includes data from 22 different robot entities (or robot types), covering a wide range of skills and tasks, totaling 150,000 tasks, including over 500 distinguishable skills.

General-purpose robots have been widely applied. The automotive industry, which uses robots the most, does not require humanoid robots but rather general-purpose robots. "Robots are moving beyond the era of large, fixed hosts," said Benjamin Gibbs, CEO of Ready Robotics.

In contrast, specialized robots can only perform a specific task. According to a research paper published in recent years comparing specialized robots and general-purpose robots, people trust the former more because robots are trained for specific tasks, reducing the chance of errors.

The future of humanoid robots

For a long time, people have been working on building humanoid robots. However, humanoid robots do not have large-scale use cases. The reasons are not hard to understand because they are very expensive to build and maintain. Secondly, they are complex and difficult to control, and finally, they are not suitable for many real-world tasks.

Although there is now a new batch of humanoid robots like Unitree, SanctuaryAI, and Fourier Intelligence, their adoption rates are low. To improve this situation, Agility Robotics is building a factory called RoboFab for large-scale manufacturing of humanoid robots. AI robotics startup Figure has released early concept images of its humanoid robot, temporarily named Figure 01.

However, compared to general-purpose robots, the adoption rate of humanoid robots is low. People are also concerned that the "humanization" of machines may harm human employment opportunities.

Some people believe that humanoid robots have potential in medical and educational applications, enhancing our understanding of robots and human beings themselves. Brett Adcock, the founder of Figure (AI Robotics), envisions humanoid robots playing various roles in fields such as enterprise labor, home assistants, and space exploration. But undoubtedly, general-purpose robots are ready to play a role in practical applications.