Nvidia Unveils the Blueprint for Training Humanoid Robots—The Next AI Revolution

Insights from Nvidia’s Rev Lebaredian: Training AI for Humanoid Robots

Meeting with Euronews Next in Taiwan

Rev Lebaredian, a leading engineer at Nvidia, recently joined the Euronews Next segment in Taiwan to provide an in‑depth explanation of how artificial intelligence can be applied to train humanoid robots.

Key Takeaways from the Interview

• AI Simulation: The approach relies on advanced simulations to condition the robot’s decision‑making processes before real‑world deployment.
• Data‑Driven Learning: Extensive datasets are used to refine the robot’s sensory perception and motor coordination.
• Real‑Time Adaptation: Algorithms enable the robot to adjust its movements dynamically in response to changing environments.
• Scalability: Nvidia’s framework supports scaling from single units to synchronized multi‑robot teams for industrial applications.

Future Directions

Lebaredian highlighted ongoing research to improve energy efficiency and reduce training time, aiming to make humanoid robots more accessible to manufacturers and researchers worldwide.

This interview marks a significant step toward integrating AI-powered humanoids into everyday technology and industry solutions.

AI‑Powered Robotics Promises to Bridge the Global Labour Gap

Nvidia’s Vice President of Omniverse, Rev Lebaredian, claimed at the Computex fair in Taiwan that the next wave of artificial intelligence involves robots. According to him, these machines could be the key to alleviating worldwide workforce shortages.

Why Robots Are the New Frontier

• Long‑standing Vision – Robotics has captivate imaginations and science‑fiction writers for decades.
• Persisting Challenge – While building the physical chassis is now routine, programming robots to act reliably remains difficult.
• General‑Purpose Breakthrough – GPT‑style AI now enables everyday users to program robots without specialized engineering knowledge.

Industry Movements

Companies such as Tesla are stepping into the race. Elon Musk’s team recently announced that its Optimus robot has begun learning household chores. Yet they concede that robots still have many lessons to learn.

Simulation: The Only Path to Intelligence?

Lebaredian stresses that robots must first learn tasks in virtual worlds for safety and efficiency. Training physically would take too long, and real‑world trials expose costly hazards:

• Data Hungry AI – Artificial systems require vast amounts of high‑quality data.
• Simulation Advantage – A digital “factory” feeds robots thousands of experiences, mirroring the data richness large language models enjoy from online resources.

Bottom Line

As AI evolves, robots will gradually become programmable by anyone, not just niche specialists. Through simulation and massive datasets, the robotics field may finally deliver the intelligence that can truly supplement, and maybe replace, the labor shortages that many countries face.

Training your robot with data

Rev Lebaredian on Generating Data for Physical AI

Rev Lebaredian, Vice President of Omniverse and Simulation Technology at Nvidia, explains why the creation of reliable training data for robots is a significant challenge. “Physical AI simply does not have an abundance of real‑world data that can be mined,” he says. “In order to teach a robot how to pick up an object we need to generate that data somehow,” he adds.

Why Real‑World Collection Falls Short

• Data acquisition from the real world is often impossible.
• Even if achievable, it can be hazardous, time‑consuming, and costly.
• Collecting enough samples across varied scenarios is practically unfeasible.

Transitioning from Fossil to Renewable Data Sources

Lebaredian argues that we must move from static “fossil” datasets to renewable data streams. The optimal renewable source for physical data, he stresses, is a physical simulator.

Industrial Use: The First Frontier

He predicts that industrial applications will be the inaugural domain to deploy these advances. “Even if we develop a perfect home robot, it’s unclear whether the general public will want one,” he notes.

Lebaredian’s Vision for Robot Learning

“A newly minted engineer starts with a vast corpus of publicly available information—textbooks, open datasets, and other common resources. Initially, this generic foundation is useful, but it only becomes truly valuable after years of targeted training on proprietary, company‑specific data.”

• General‑purpose training provides a baseline.
• Specialized corporate data tailors the robot to specific workflows.
• Dedicated training ensures optimal performance within the organization.

Future Outlook

While he hasn’t given a precise timeline, Lebaredian is optimistic about the near‑future appearance of humanoid robots. “It’s coming soon,” he concludes.

For more on Nvidia’s latest developments, see the related coverage:
  ‘Computers of the future,’ Disney robots and AI factories: The key takeaways from Nvidia’s GTC 2025.

Where and what to use your robot for

Robots Are Poised to Transform Production and Warehousing

Industry is likely to adopt robotic assistance first. In contrast to consumer homes, factories and warehouses represent the immediate frontier for deploying advanced automation.

Key Challenges Fueling the Need for Machines

• National retirement rates are eroding the workforce of experienced technicians.
• The OECD reports that labour deficits are at the highest point seen in a decade.
• Digital, repetitive, and hazardous duties—known as “three‑D jobs”—are increasingly unattractive to younger workers.

Why Taiwanese Policy is Contingent on Robotics

The government has announced a five‑year agenda to strengthen its robotics sector, a move intended to cushion the impact of a shrinking population on economic resilience and caregiving infrastructure.

Peter Hong, leader of the National Science and Technology Council’s Engineering & Technologies division, said the country’s future prosperity hinges on integrating new machine platforms into the industrial chain.

Beyond the Factory Floor: A Vision for Humanoid Robots

According to industry expert Lebaredian, the evolution of robotics will expand into retail environments—helping retailers manage shelf stocking when human labor is scarce. He also cited potential applications in mining, nuclear power plants, and even space missions.

Ultimately, the technology could serve as a companion for elderly care, offering assistance where demand grows.

Related Topics

• Will autonomous robots replace maritime crews or open new career pathways?
• What role will Nvidia’s AI supercomputing initiatives play in supporting Taiwan’s technological ambitions?

How to make your robot safe

Balancing AI’s Ambition with Its Real‑World Perils

As the next stage of artificial intelligence unfolds, the excitement is palpable—but so is the caution. Large language models, like ChatGPT, still generate inaccuracies, and when such errors translate into the physical world, the danger escalates.

Improvements Across Two Years and a Half

Key Insight: The accuracy and quality of generative AI have surged dramatically since the launch of ChatGPT. Yet, even with rapid advancements, complete precision remains elusive.

The Human Element in AI Tasks

“There isn’t ever a single ‘right answer’ when the tasks require human-like judgment,” noted Lebaredian. This uncertainty underscores the necessity for ongoing refinement.

Measurable Industry Applications

In sectors where outcomes are quantifiable—such as:

• Grasping an object accurately
• Transporting it safely to a new location
• Executing the task reliably under varied conditions

—robust testing protocols can ensure performance before deployment.

Safety Lessons from Nuclear Technology

Just as nuclear reactors and other high‑risk systems undergo stringent safeguards, physical AI must adopt parallel safety standards:

• Rigorous design validation
• Comprehensive simulation testing
• Continuous monitoring during operation

Looking Ahead

While autonomous vehicles once seemed like science fiction, society has adapted as the technology matured. The same trajectory is expected for AI, with iterative upgrades enhancing reliability and safety.

Economical Implications for Europe

Europe is evaluating how to keep pace with AI hardware leaders like Nvidia—an endeavor that plays into broader discussions about AI infrastructure and competition.

Share this transformed perspective with colleagues and keep the dialogue about AI’s future vibrant and responsible.