Humanoid Robots Face Training Gap as Industry Demands Practical Applications Over Demos

2 Sources

Share

At CES 2025, humanoid robots showcased impressive feats like dancing and playing ping-pong, but industry insiders are pushing for real-world utility. Experts warn that achieving true autonomy will take years, as current AI struggles with embodied learning. The robotics industry projects a $179 billion market by 2030, with factories leading adoption.

News article

Humanoid Robots Steal CES Spotlight But Questions Remain

At the Consumer Electronics Show (CES) this week, humanoid robots captivated audiences with their ability to dance, somersault, deal blackjack, and play ping-pong. Yet behind the spectacle, industry experts voiced growing impatience for these machines to move beyond entertainment and deliver genuine utility

1

. Christian Rokseth, founder of Humanoid Guide, captured the sentiment bluntly: "They've shown robots dancing and doing kung fu; now show us that they can be productive"

2

. While the hardware side has accelerated over the past year, the path to practical robots capable of operating in kitchens or on factory floors remains fraught with challenges.

AI and Training Limitations Block True Autonomy

The core obstacle preventing AI-powered humanoid robots from achieving true autonomy lies in how they learn. To become autonomous, humanoid robots need AI that translates sensory input into physical actions—a capability that extends far beyond today's large language models powering tools like ChatGPT

1

. Henny Admoni, an associate professor at Carnegie Mellon University's robotics institute, explained the fundamental problem: "If you want (robots) to learn embodied things, you have to put them inside a body"

2

. Training large language models relies on massive internet data, which proves largely useless for robots that need to navigate physical spaces and manipulate objects. Rokseth likened the current approach to locking a child in a room and expecting it to learn about the world—embodied learning requires real-world interaction that can't be simulated through text alone

1

.

Tech Giants Partner to Build Robot Brains

Despite these training limitations, companies are forging ahead with ambitious partnerships. EngineAI founder Evan Yao revealed that the Shenzhen-based company collaborates with tech titans including Amazon and Meta to develop AI brains for its creations

2

. "We are trying to simulate humans, but the robots will never become human," Yao told AFP, acknowledging the emotional and cognitive gaps that will persist. Meanwhile, South Korean automotive giant Hyundai unveiled Atlas, a humanoid robot created with Boston Dynamics, which it plans to test in factories

1

. The robotics industry shows clear dynamism, with the Consumer Technology Association projecting the global market will hit $179 billion by 2030

2

.

Factories May Lead the Way for Industrial Applications

The bulk of market growth is expected in factories, warehouses, and other business operations where robots work in controlled environments

1

. Artem Sokolov, founder of a London-based humanoid robotics startup, argues that since humans already work in factories, robots copying their bodies can thrive on assembly lines too. This focus on industrial applications makes strategic sense given current technological constraints. However, Admoni cautioned that many companies claiming to develop autonomous humanoid robots are actually using teleoperation—where a person in a suit or using controllers translates every movement to the robot

2

. This reality check matters for investors and businesses evaluating which robotics ventures can deliver near-term value versus long-term promises.

New Training Methods Target Uncontrolled Environments

To overcome the embodied learning challenge, startups are experimenting with novel training approaches. Some companies now have people wear cameras and haptic gloves while performing chores at home, capturing real-world movement data that robots can learn from

1

. Rokseth emphasized that "to make robots general machines, they need to be let out in the real world," not confined to assembly lines or warehouses. This shift toward training in uncontrolled environments represents a critical evolution for the robotics industry. Meanwhile, Robotera showcased a humanoid robot training to complete the Beijing marathon in coming months—a demonstration that physical endurance tasks may arrive before complex manipulation skills

2

. For businesses watching this space, the timeline for truly capable practical robots appears to stretch several more years, requiring patience alongside investment in foundational AI research that bridges the gap between language understanding and physical action.

Today's Top Stories

TheOutpost.ai

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

© 2026 Triveous Technologies Private Limited
Instagram logo
LinkedIn logo