Ask Onix
Humanoid robots trained with children's blocks
A shiny, faceless robot named Phantom practices stacking colored blocks in a San Francisco industrial space, far from the battlefield imagery often associated with military robotics. The scene is part of Foundation Robotics' effort to gather data on how its humanoid machine interacts with its surroundings.
From supply runs to frontline combat
Foundation Robotics, a two-year-old startup, is developing Phantom for both military and civilian applications. Co-founder and CEO Sankaet Pathak outlines a broad range of potential roles, including logistics, reconnaissance, casualty recovery, and hazard inspection. The company also envisions the robots engaging in combat to neutralize threats-a concept Pathak describes as "frontline weaponization."
Pathak argues that armed robots could reduce risks to human soldiers by navigating dangerous environments like urban combat zones. He also claims that ground-based autonomous systems could minimize collateral damage compared to aerial strikes. However, Phantom MK-1, the first-generation model on display, lacks key capabilities: it has no battery, is not weatherproof, and cannot recover if it falls. Its hands are weak, and it lacks functional wrists.
Next-gen upgrades and mass production goals
A second-generation model, Phantom MK-2, is under development in a restricted area of the facility. Pathak says it will feature weatherproofing, a six-hour battery life, and the ability to self-right after a fall. Improved hands with functional wrists are also in the works, designed to handle weapons. The company aims to produce at least 40,000 units annually by the end of 2027, with a long-term target price below $20,000 per unit.
Pathak frames the project as a response to China's advancements in robotics, warning that the West risks falling behind. He envisions hundreds of thousands of AI-driven humanoid robots forming a ground force to match the growing use of autonomous drones. Foundation has secured $24 million in research contracts with the U.S. military and is testing two units with the Ukrainian military. While the U.S. pilot focuses on non-combat tasks, Ukraine's testing includes weaponization.
Military interest and ethical debates
The U.S. military's interest in humanoid robots is evident in initiatives like an ongoing Army contest for multi-role humanoid systems. Dean Fankhauser, an advisor at Robozaps, calls the weaponization of such technology "completely inevitable," noting that simpler robots, including drones and quadrupeds, are already used in combat. However, some companies refuse to weaponize their robots, citing ethical concerns.
Pathak dismisses these objections, arguing that humanoid robots are practical because they can use existing human tools, from screwdrivers to firearms. He advocates for human oversight in lethal decisions but acknowledges scenarios where autonomous action might be necessary to prevent catastrophic outcomes.
Technical hurdles and real-world challenges
The biggest obstacle for humanoid robots is developing AI capable of navigating unpredictable environments. Phantom's AI system, Cortex, combines a "reasoning model" trained on task-specific data with a "world model" that predicts environmental responses using internet videos and real-world interaction data. The robot relies on 360-degree cameras for navigation and adapts its movements based on its surroundings.
Yet skeptics question whether the humanoid form is the most effective. Fankhauser points out that quadruped robots can navigate terrain more efficiently. He also notes that current commercial humanoids struggle with basic tasks like opening doors, let alone complex military operations. "If there was a war in Taiwan today, the idea that China could deploy militarized humanoids effectively is fanciful," he says.
Robert Griffin, who works on humanoid robots at the Florida Institute for Human and Machine Cognition, highlights the difficulty of operating in unpredictable environments. Tasks like jumping through windows or navigating unfamiliar interiors remain significant challenges. Griffin also notes that human soldiers have easily foiled AI systems by behaving unpredictably, such as somersaulting or covering their heads with boxes.
Ethical concerns and the future of warfare
Nicole van Rooijen, executive director of Stop Killer Robots, warns that lethal autonomous weapons lower the threshold for conflict and dehumanize warfare. She finds the humanoid form particularly concerning, as it may appear trustworthy, increasing the risk of misjudging danger. Van Rooijen advocates for international regulations to curb the arms race in autonomous weapons.
Despite the challenges, Foundation Robotics is pushing forward. Pathak acknowledges the technical difficulties but remains optimistic. "We're setting extremely hard challenges for our engineering team," he says, "and we'll either meet them or fail."
