For decades, the popular imagination painted Japan as the undisputed “robot kingdom.” Visions of sophisticated androids assisting in daily life, born from a culture steeped in manga and technological fascination, seemed perpetually just around the corner. South Korea, too, with its tech-savvy populace and industrial giants, appeared poised to be a major player in the burgeoning field of robotics. Yet, as a global humanoid robot revolution ignites, fueled by advancements in artificial intelligence and a growing demand for automation, a surprising narrative is unfolding: China is surging to the forefront, leaving its East Asian neighbors struggling to keep pace.
Recent industry analysis paints a stark picture. A Morgan Stanley research report, aptly titled “Humanoid 100,” meticulously cataloged the top 100 publicly listed companies driving the global humanoid robot supply chain. The geographical distribution of these key players was revealing: China and North America (US and Canada) each accounted for an impressive 35 companies. The rest of the world lagged significantly, with the broader Asia-Pacific region (excluding China) claiming a mere 18 spots, and Europe, the Middle East, and Africa combined contributing just 12.
Strikingly, Japan, the nation synonymous with robotics innovation, was virtually “missing in action.” Bundled into the “Asia-Pacific” category alongside South Korea, Japan’s once-dominant presence in the global robotics landscape seems to have faded into the background. This absence isn’t limited to publicly traded companies; even within the ranks of private humanoid robot unicorns, Japan is conspicuously absent.
This apparent eclipse is particularly perplexing considering Japan’s pioneering legacy in humanoid robotics. As early as 1973, Japan’s Waseda University unveiled WABOT-1, a groundbreaking creation recognized as the world’s first full-scale humanoid robot. Beyond being an early starter, Japan also boasts a significant number of robotics companies, reportedly ranking third globally in 2024 with 26 firms, trailing only China (43) and the United States (28).
However, amidst the current wave of “embodied intelligence,” driven by breakthroughs in artificial intelligence, Japan’s humanoid robot sector seems to have fallen silent. No groundbreaking, impactful products have emerged from the nation that once dreamt of becoming a “robot kingdom.” Japan, it seems, got an early start but missed the crucial turning point.
The Fading Dream of a “Robot Kingdom”: Japan’s Half-Century Head Start
Japan’s journey in humanoid robotics began with an undeniable lead. WABOT-1, born in the labs of Waseda University in 1973, wasn’t just a machine; it was a symbolic leap. While rudimentary by today’s standards, WABOT-1 possessed rudimentary vision, hearing, and basic hand-eye coordination. It could grasp objects and exhibited a level of “intelligence” comparable to an 18-month-old child. Its appearance, however, was decidedly utilitarian – a frame seemingly cobbled together from salvaged materials, reminiscent of a post-apocalyptic contraption.
The subsequent WABOT-2, unveiled shortly after, began to embody a more “human” aesthetic. Crafted from polished black metal, carbon fiber reinforced plastic, and intricate wiring, WABOT-2 could perform more sophisticated tasks. Its “tubular legs” could operate pedals, and its “steel fingers” could caress piano keys, even capable of playing Bach’s “G Major String Suite.” Legend has it that the then-aged Japanese Emperor was deeply captivated upon witnessing WABOT-2’s performance.
During the same era, while Western nations, particularly Germany and the United States, were making strides in industrial robotics and artificial intelligence – KUKA in Germany launched the world’s first six-axis industrial robot FAMULUS, and the Stanford Research Institute in the US developed Shakey, the first intelligent mobile robot – they lagged behind Japan in the pursuit of “humanizing” robots.
Japan’s early dominance in humanoid robotics was a confluence of factors. A key driver was the Japanese government’s unwavering commitment to establishing a “robot kingdom,” coupled with a deep-seated cultural affinity for humanoid robots among the populace. The 1960s and 70s witnessed a growing labor shortage in Japan. A 1965 report by the Japanese Ministry of Labor highlighted a need for 1.8 million skilled workers. Companies like Nissan even resorted to employing seasonal and farm workers to address the manpower gap. The 1973 oil crisis exacerbated the situation, driving up production costs and overall living expenses, leading to demands for higher wages and further inflating labor costs.
To alleviate the strain on the workforce, Japan embraced the “robot nation” strategy, envisioning automation driven by robotics technology as the solution. Simultaneously, Japanese society displayed a remarkable receptiveness to humanoid robots, fostered in part by the influence of iconic robot anime. In 1963, Osamu Tezuka’s “Astro Boy” deeply inspired a generation of young researchers. University students participating in bipedal robot debates proudly proclaimed themselves “Astro Boy’s generation,” finding the prospect of “creating something that could walk” inherently fascinating. Similarly, the 1979 anime classic “Mobile Suit Gundam,” featuring human-piloted humanoid mecha, captivated Japanese youth.
For this generation, humanoid robots were not menacing automatons but rather companions, capable of coexisting harmoniously with humans, offering assistance rather than posing a threat. Against this backdrop, Japan poured resources into robotics research, seemingly without budgetary constraints. Humanoid robots became symbolic artifacts in Japan’s pursuit of its great power aspirations.
Japanese automotive giant Honda emerged as a frontrunner. Embarking on humanoid robot research in the 1980s, Honda iteratively developed the P1, P2, P3, and P4 prototypes. Each iteration brought advancements in lightness, gait stability, and the integration of robotic arms, culminating in the year 2000 with the unveiling of the globally acclaimed ASIMO.
Named after science fiction author Isaac Asimov, ASIMO was heralded as the world’s first humanoid robot capable of interacting with humans in a truly human-like manner. It was also the first bipedal robot powered by batteries capable of navigating inclines. ASIMO’s intended role was to assist humans, potentially aiding the elderly or performing community tasks. ASIMO became the embodiment of harmonious human-robot interaction, performing at exhibitions and even “receiving” dignitaries. In 2014, then-US President Barack Obama famously kicked a soccer ball with ASIMO during a visit to Japan. Remarkably, at this time, Boston Dynamics, now considered one of the world’s most advanced robotics companies, had not yet ventured into humanoid robot development.
Following ASIMO’s debut, humanoid robot projects proliferated in Japan. Beyond major corporations, research institutions and universities joined the fray. However, few of these initiatives yielded truly standout products. The National Institute of Advanced Industrial Science and Technology (AIST) and Kawasaki Heavy Industries collaborated on the HRP series of humanoid robots, aiming to explore applications in healthcare, industry, and disaster relief.
This period also saw an increasing emphasis on anthropomorphism – making robots as human-like as possible. The University of Tokyo and Kitano Symbiotic Systems jointly developed Pino, an open-source robot, to study human-like robotics technology. Hiroshi Ishiguro, a robotics expert and engineer from Osaka University, created 1:1 scale replicas of humans, capable of facial expressions and interaction through algorithms, sparking discussions about the “uncanny valley” effect.
In 2014, Japanese capital began to play a more direct role. SoftBank, under the leadership of Masayoshi Son, strategically acquired two key robotics ventures. One was Pepper, the second most recognized Japanese humanoid robot after ASIMO. However, Pepper was originally developed by French company Aldebaran Robotics, which SoftBank acquired. Pepper was marketed as the first personal humanoid robot for home use, capable of recognizing and responding to human emotions, engaging in conversations, and performing basic interactive tasks like greetings and information provision.
Building upon Pepper’s technological foundation, SoftBank acquired Boston Dynamics from Google in 2017. At the time, Boston Dynamics’ Atlas prototype, unlike Pepper, focused less on human interaction and more on demonstrating advanced locomotion in complex environments. However, SoftBank’s ownership of Boston Dynamics was short-lived. Just three years later, in 2020, SoftBank sold Boston Dynamics to South Korean automotive giant Hyundai Motor Group. This marked a significant retreat of Japanese capital from the forefront of humanoid robotics development.
Looking back over half a century, Japan had relentlessly pursued a grand vision of a “robot kingdom,” but this dream proved to be extravagantly expensive. The cost of a single WABOT already reached $1 million USD. Compounding the financial burden, humanoid robots lacked readily available, commercially viable applications. When the massive investments failed to generate commensurate returns, the dream began to fade.
The Wake-Up Call: When the Humanoid Robot Dream Faltered
Humanoid robots, after all, are not truly human. They shouldn’t be subject to the cycle of birth, aging, and death. Yet, Japan, after a half-century sprint in this field, seemed to be entering a phase of “fifty and knowing destiny” – a sense of resignation.
This “knowing destiny” manifested as a lack of momentum. After 2010, Japan’s investment in humanoid robots ceased to be “unlimited.” Maintenance of existing achievements became the priority, with limited innovation and reduced financial backing. This shift stemmed from the premature entry into humanoid robotics before the market was ready. Initial investments failed to yield expected returns. It was as if Japan had exerted too much force too early, exhausting its energy reserves.
From 2018 onwards, prominent Japanese humanoid robot projects began to be discontinued or divested. Honda’s ASIMO, whose core purpose was to assist the elderly and those in need, ceased production in 2018. With a staggering cost of $2.5 million per unit, ASIMO was far more expensive than employing a human caregiver. It became more of a “pet project” within Honda, rather than a profitable business venture.
SoftBank, known for its bold investments, followed suit in 2020, halting production of its star robot, Pepper. Unlike ASIMO, which lacked commercial application, Pepper, despite lukewarm demand, had found niche deployments in commercial settings. Pepper’s relatively low price point – around 198,000 yen (approximately $1,000 USD at 2020 exchange rates) – made it more accessible. SoftBank itself deployed a large number of Peppers in its retail stores. During the pandemic, SoftBank even used 100 Peppers as cheerleaders for its “Hawks” baseball team.
However, Pepper’s performance was often inconsistent, leading to frequent “firings.” Peppers were known to abruptly stop and “perform” inappropriately during funerals or consistently misdirect customers to the “liquor section” regardless of their inquiries. In the same year Pepper production ceased, SoftBank sold 80% of Boston Dynamics to Hyundai Motor Group for 1 trillion won ($917 million USD), further signaling its retreat from humanoid robotics.
Both Boston Dynamics and Aldebaran Robotics (Pepper’s developer) were undeniably leading companies in humanoid robotics. However, they shared a common challenge: their products were either prohibitively expensive or lacked sufficient practicality, resulting in sluggish sales. In response, both Honda and SoftBank chose to discontinue production rather than pursue further iterations.
Beyond the inherent commercial challenges of humanoid robots, Japan’s decision to pull back was also influenced by broader economic realities. In March 2011, a magnitude 9.0 earthquake and tsunami devastated northeastern Japan, causing unprecedented damage and plunging the Japanese economy into prolonged stagnation. The International Monetary Fund (IMF) reported that Japan’s average annual growth rate between 2010 and 2020 hovered around 0.8%-1.0%, significantly lower than the global average.
Simultaneously, Japanese companies that had championed humanoid robotics faced their own financial headwinds. Honda’s global auto sales growth was declining. While discontinuing ASIMO in 2018, a non-profit venture, Honda’s sales decline continued, dropping 10.01% year-on-year in 2019.
SoftBank, too, faced a liquidity crunch. The failed IPO of WeWork, a heavily invested SoftBank venture, plunged WeWork into a severe financial crisis, teetering on the brink of bankruptcy. SoftBank’s total investment and support for WeWork amounted to $19.5 billion USD. The IPO debacle forced SoftBank to reassess its business strategy, prompting its withdrawal from capital-intensive, low-commercialization ventures like humanoid robotics.
Following these production halts and divestments, Japan’s humanoid robotics sector, once vibrant for half a century, entered a period of dormancy. Japanese humanoid robotics entered a silent phase. But as Japan retreated, the rest of the world, particularly China and the United States, experienced a surge in humanoid robot activity. In 2022, Elon Musk announced Tesla’s entry into humanoid robotics at Tesla AI Day. In 2023, Figure AI’s collaboration with OpenAI propelled humanoid robot intelligence to new heights. In 2024, Chinese humanoid robot companies like Unitree Robotics, Fourier Intelligence, and Agile Dynamics secured substantial funding rounds.
This stark contrast – Japan’s quiescence and the global resurgence – highlights a critical divergence. Japan and the rest of the world were perfectly out of sync in the unfolding humanoid robot revolution.
The AI Chasm: Where Japan Missed the Boat
The 2020s mark a clear turning point, witnessing the humanoid robot momentum shift from Japan to other global players. Before 2020, Japan spearheaded the first wave of humanoid robot development. After 2020, China and the United States are leading the second wave.
To the casual observer, the capabilities of robots from both eras might appear superficially similar. Today’s humanoid robots can walk bipedally and perform tasks like serving tea, functions seemingly within ASIMO’s repertoire. ASIMO even demonstrated the ability to effortlessly open a bottle cap on a television program.
However, these seemingly identical actions are underpinned by fundamentally different technologies. In the first phase, robots like ASIMO relied on traditional motion control algorithms. They could execute pre-programmed actions but had limited adaptability to changing environments. In contrast, today’s humanoid robots leverage AI large models and reinforcement learning. This enables a degree of scene generalization and autonomous decision-making, allowing for greater environmental adaptability. In essence, earlier robots were like marionettes, limited to predefined movements. Modern humanoid robots possess a “brain,” capable of executing tasks independently.
The pivotal shift from the first to the second wave of humanoid robotics lies in advancements in artificial intelligence – precisely where Japan faltered. From the dawn of the internet era to the rise of cloud computing, Japan has consistently lagged behind the global curve in related industries. This lag in internet and cloud computing infrastructure further hampered the development of AI technology in Japan, which heavily relies on these foundations.
In the internet era, China birthed tech giants like BAT (Baidu, Alibaba, Tencent), and the US saw the rise of Google, Amazon, and Facebook (now Meta). Japan, in contrast, lacked comparable internet behemoths. This deficiency translated into a disadvantage in internet data accumulation, a crucial resource for AI development.
In the AI era, starting from late 2022 and early 2023, AI large models, spearheaded by OpenAI, swept the globe. US tech giants Google, Meta, Anthropic, and Elon Musk’s X swiftly followed suit. China also witnessed the emergence of unicorns like DeepSeek, with Baidu, Alibaba, Tencent, Zhipu AI, MiniMax, and others rapidly entering the arena, deploying AI in applications ranging from humanoid robots to social media and autonomous vehicles.
Japan, however, found itself relegated to the role of “follower” in the AI race. The nation faced a scarcity of domestic AI talent. Top-tier research institutions like RIKEN relied heavily on foreign scholars, particularly from China, for research papers. Large model development heavily depended on foreign capital. For instance, SoftBank partnered with OpenAI to establish “SB OpenAI Japan,” relying on US-manufactured chips for training large models.
Constrained by these factors, Japan struggled to keep pace with global AI advancements. Yet, Japan had, at one point, briefly held a leading position in AI research. In 1988, Japanese corporation NEC established its Artificial Intelligence Research Institute, NEC Lab, in Princeton and Silicon Valley in the US. NEC Lab attracted top AI researchers globally, earning the moniker “the West Point of AI.” Yann LeCun, a pioneer of convolutional neural networks, was once a researcher at NEC Lab. Kai Yu and Yuanqing Lin both served as directors of NEC Lab, which also nurtured talent that later contributed to China’s BAT companies.
However, Japan, at the time, dismissed “deep learning” as an outdated technology. The nation doubled down on the “symbolism” technology route, investing hundreds of millions of dollars in the “Fifth Generation Computer” project, focusing on expert systems based on logical editing, aiming to achieve intelligence through manually inputting knowledge bases.
This misguided approach led to a talent drain from NEC Lab, as researchers advocating for deep learning gradually departed. In 1992, the Fifth Generation Computer project, after consuming hundreds of millions, was declared a failure. This setback severely dampened Japan’s confidence in AI research. AI research became a pariah in Japanese academia, and funding for neural network research dwindled. Even after AlphaGo’s victory over Lee Sedol in 2016, Japan remained skeptical of deep learning. Some academics even commented, “If a report title contains ‘deep learning,’ the audience might be sparse.”
Overall, whether considering humanoid robots or AI, Japan once held a global lead, but these advancements seemed to be untimely. A combination of historical factors, compounded by Japan’s prolonged economic downturn and a traditional emphasis on hardware, caused them to miss the global technological shift.
Conclusion: A Marathon Run Astray
With the advent of AI large models, Japan’s humanoid robot industry has shown faint signs of renewed interest. Efforts have been made to integrate ChatGPT into Pepper robots to enhance their emotional expressiveness. In late 2024, Osaka University developed “waveform motion” dynamic facial expression synthesis technology, overlaying facial action waves like blinks and frowns to generate natural expressions, addressing the problem of robotic facial stiffness.
These developments might suggest a resurgence in Japanese humanoid robotics, but closer inspection reveals that these efforts remain largely superficial improvements. Masayoshi Son’s greatest regrets include the WeWork investment failure and his lament over Japan’s technological lag, for which he reportedly wept.
The history of Japanese humanoid robotics is poignant, akin to a marathon runner who sprints ahead with all their might at the starting line, only to be overtaken by competitors halfway through, having exhausted their energy and unable to muster a final surge. Sometimes, being born at the wrong time is a profound tragedy. As once-promising robots are shelved, forgotten, and abandoned, the former glory fades into a sigh of bygone eras.
In stark contrast to the hesitant steps of Japan and the industrial focus of South Korea, China is charging ahead, leveraging its unparalleled manufacturing prowess and rapidly developing AI capabilities to build a comprehensive humanoid robot industry. While challenges remain, particularly in AI model development, China’s robust supply chain, combined with a dynamic and competitive tech ecosystem, positions it as a formidable contender, poised to potentially redefine the global landscape of humanoid robotics. The “robot kingdom” dream, once envisioned for Japan, may very well be taking root in the land of the dragon.
评论