Jan Liphardt, CEO and founder of the US robotics operating system (OS) startup ‘OpenMind’, stated this in an interview held on the 23rd of last month in Gangnam-gu, Seoul. OpenMind plans to test deploy the thinking ‘robotic dog’ in about ten households in San Francisco by the end of this month.

Founded in 2024 by Stanford University professor Jan Liphardt, ‘OpenMind’ aims to be the ‘Android’ of the robotics operating system (OS) field. OpenMind has developed the universal robotics operating system ‘OM1’, with the goal of enabling all AI-connected robots to operate on a single language and platform. Their system has been installed in the globally recognized Chinese Unitree ‘G1’ humanoid robot. OpenMind attracted an investment of USD 20 million this August, with backing from venture funds like ‘Pebblebed’, established by founding members of OpenAI.

OpenMind believes that a new robotics operating system that can ‘think’ like humans is necessary in situations where much more interaction between humans and machines is required, such as with humanoid robots in homes. A characteristic of OpenMind is its orientation towards ‘open source’. This allows software to be installed on any robot without being tied to a specific manufacturer's ecosystem or having to recreate existing technology each time. Currently, the robot safety technology market largely relies on individual robot manufacturers' proprietary solutions, leading to a high demand in the industry for a standardized safety platform. Below is a Q&A with CEO Liphardt.

When robots were deployed in nursing homes in the US where elderly people with dementia reside, something remarkable happened. All the robots did was make these people smile through conversation. However, at some point, the nursing home nurses had to wipe lipstick marks off the robots' heads every evening. This was because the elderly, who were not frequently visited by their children, felt a sense of closeness to the robots and kissed them daily. It was astonishing to see the elderly deeply interacting with the robots.

Just a few years ago, the biggest constraint on the development of robots was technological factors. But not anymore. Currently, robot hardware and software are advancing very rapidly. The barriers we face now are issues like cost, safety, and insurance. These issues need to be well resolved for the ‘one robot per household’ moment to arrive.

For example, consider deploying robots in kindergartens, schools, homes, and fire stations. How will safety regulations be established? Additionally, there are issues of legal and financial arrangements.

OpenMind is one of the few companies with robot insurance. Munich Re, the world's largest reinsurance company, has even launched a robot insurance product.

This is because questions arise, such as whether one can receive insurance money if a humanoid robot in the home steps on one's foot and causes injury. Insurance is provided when a robot fails to deliver promised services, makes poor decisions, or injures someone. These are the new hurdles facing the robot industry.

Our priority is to quickly provide robots to the elderly and children. We need to receive feedback from the field to understand what roles are truly necessary for them. People ask when robots that organize socks in drawers will be available, but such robots are less profitable and will be lower in priority. The current focus of the robotics industry is on humanoids that perform tasks in education, nursing, plumbing, and electrical work.

The robotic dogs we are deploying in San Francisco households from the end of this month are similar. They cannot perform tasks like ballet or chopping onions, but they can teach math to children, safely care for the elderly, and handle ‘security’ to prevent home intrusions. If a robotic dog does not see its owner for more than 10 minutes, it will search the house and check on the owner to ensure they are okay.



Reflecting on my experience as a professor lecturing to thousands of students at Stanford University, when people learn something, they must start from the beginning, and it is difficult to immediately pass on the knowledge they have just learned to others. It is like not being able to download Korean language skills and speak it immediately.

However, ‘machines’ like robots can model a learned skill digitally and share it immediately with other ‘machines’. For example, if a robot in a manufacturing plant learns a specific skill well, it can instantly pass that knowledge and skill to other robots. What we need to develop now is the ability for robots to learn and spread such skills and know-how effectively, but in a non-threatening way.

Unfortunately, there are none yet. Although robot technology is advancing rapidly, there are still no safety standards in place. If we call the San Francisco City Hall and ask, ‘Can a humanoid cross the street?’, no one at the city hall can answer. They simply respond, “That is not my department. Call somewhere else.”

Korea is a very fortunate country. It is rare for a country to have all the resources and infrastructure necessary to succeed in the robotics industry, and Korea is one of them. There are five universities in Seoul alone focusing on robotics research. In addition, Korea has the cutting-edge manufacturing capabilities of large corporations like Samsung Electronics and LG Electronics. Can a completely new company emerge and mass-produce robots? For large-scale robot production, accumulated manufacturing experience is essential. In this regard, Korea has the formidable manufacturing capabilities to produce millions of humanoid robots.

The core of this research is to preemptively mitigate ‘hallucinations’—dangerous actions or incorrect judgments by robots—within context. To achieve this, the multimodal (various signals such as video and audio) safety guardrail layer powered by OpenMind's robot operating system OM1 comprehensively assesses the safety of proposed actions by understanding the surrounding situation in real time and controls the final behavior.

For instance, if a robot attempts to move with a sharp object while a child is nearby, it is judged as a ‘risk of injury’ and the action is blocked. Currently, most robots are at the level of receiving simple commands for basic safety rules like Asimov's three laws of robotics. Companies have faced limitations in safely operating robots in complex real-world environments. OM1 will serve as a comprehensive platform integrating safety functions across all areas of robotics.

Popular News