Written by: Techub News Organization

Introduction

In a recent in-depth dialogue hosted by Logan Kilpatrick, Brett Adcock, founder and CEO of Figure AI, openly shared his sharp insights on the future of humanoid robots and the integration of artificial intelligence with the physical world. As a leader of a star company that has moved humanoid robots from concept to real-world deployment in just a few years, Brett Adcock's thoughts are not only based on Figure's own frontline practices but also outline the paradigm shift facing the entire industry. This dialogue is critical as it reveals how humanoid robots have leaped from a long-stagnant "sci-fi concept" to being the most promising physical carriers of AI, which may reshape the labor market and home life in the next decade.

Summary

• The reliability of humanoid robot hardware has undergone a qualitative leap, evolving from dangerous, short-lived hydraulic systems to safe, durable, and scalable electric drive systems.
• The core of robotic learning has shifted to "neural network issues," with end-to-end training models enabling robots to perform complex tasks in ways close to human capabilities, something that was not possible five to ten years ago.
• Figure AI's strategy is to directly tackle the "ultimate form" of humanoid robots, avoiding compromises in design to escape local optima, and choosing to enter structured labor markets (logistics, manufacturing) to accumulate capabilities and data for eventual entry into challenging household environments.
• Humanoid robots are seen as the ultimate deployment carrier of AGI (Artificial General Intelligence) in the physical world, and in the coming years, their physical deployment numbers will begin to grow exponentially, eventually reaching a scale comparable to that of humans.
• The default interface for robot interaction will be natural language (voice and text), with privacy, security, and large-scale integration being one of the most pressing challenges currently faced.

Hardware and AI: The Dual Engines Driving Exponential Progress in Humanoid Robots

Brett Adcock pointed out straightforwardly that the current "exponential" development in the humanoid robot field is primarily due to two fundamental changes: the significant improvement in hardware reliability and the successful application of neural networks in robotics.

First is hardware. Adcock compared the difference between ten years ago and now: "The best robot ten years ago might have been Boston Dynamics' Atlas, which was a hydraulic system, could only run for about 20 minutes, leaked oil everywhere, and operated at pressures of thousands of PSI, extremely dangerous for humans." Now, companies like Figure have been able to build "extremely reliable" electric humanoid robot hardware systems. He confidently stated that if one visits their factory, they would see robots everywhere, and "one might not see a hardware failure all day long." This reliability is the prerequisite for achieving large-scale deployment, akin to the fact that without reliable basic hardware, no advanced AI would have a role to play. He likened this complexity to that of rockets and turbofan engines but firmly believes humanoid robots can achieve similar industrial-grade reliability.

Second is AI and learning capability. Adcock believes that the robotic problem is "essentially a neural network problem." It is only in recent years that neural networks have truly begun to exert their power in robotics. The Helix neural network running on Figure's robots can command the robot's hand positions, head, torso, and the entire action space in an end-to-end manner. He mentioned a video released a week ago: the robot performed logistics work continuously for 60 minutes, all controlled by a single, locally operating S1 Helix neural network. "Its behavior is very much like that of humans... we are approaching human speed and performance." This ability to learn driven by data, mimicking human behavioral patterns, was completely non-existent five to ten years ago.

These two major advancements—reliable bodies and learning brains—constitute the foundation of the humanoid robot explosion. Adcock emphasized that only when the two are combined can the potential of robots to accomplish useful work in the physical world be unlocked.

Why Must It Be Humanoid? A Strategic Path from Labor Market to Home

Faced with the question of "why not first create simpler wheeled or specialized robots," Brett Adcock's response was remarkably firm: it is essential to directly tackle humanoid robots. On the first day of founding Figure, he wrote and publicly released a general plan to directly create humanoid robots.

His logic is that choosing other forms (like wheels or claws) is a "local optimal solution" that becomes a significant "crutch" for business development, ultimately hindering one from achieving the true goal—creating a general robot that can work in modern human environments like humans do. He has believed for three years that directly creating humanoid robots is feasible, thus the company has focused on the design and construction of humanoid robots from day one.

Regarding market entry paths, Figure has two main focuses: the labor market and the home. Interestingly, although the home market has enormous potential, Adcock admits that "the home is much harder than the labor market." The reason is that "the engineering challenges are proportional to the variability of the environment." Workplaces (logistics, manufacturing, medical, construction) are usually more structured and less variable, making it easier to integrate autonomous systems. Each household, however, presents a unique, highly unstructured environment with immense challenges.

Therefore, Figure adopts a pragmatic "two-step" strategy: first "kick-starting" in the labor market to validate technology, accumulate data, and achieve commercialization; while simultaneously vigorously advancing research and development in home scenarios. Adcock predicts that the maturity of home scenarios will occur in "single-digit years (within a few years)," at which point robots will be able to genuinely perform useful autonomous work for humans.

He uses logistics collaboration as an example to illustrate their flexible strategy: when signing a contract with a logistics company, they did not yet know what exactly to do. After field inspections, they identified small package sorting as a pain point and opportunity. This task is highly challenging—each package is different, and the stacking methods constantly change, making it unsolvable through traditional coding. This prompted them to fully engage in the learning pathway, and as a result, they quickly saw success. He revealed that the sorting speed of the robot increased from "one package every four seconds" to "one package every three and a half seconds," and it could perform more intelligent operations, such as moving one package to process another, demonstrating a reasoning ability similar to a "chain of thought."

The ultimate goal is to create a general robot platform: theoretically, as long as the robot is equipped with the appropriate range of motion, load, and speed, it should be able to complete most of the work in its facility that humans do.

Challenges of Scaling: Manufacturing, Learning, and Coexisting with Humans

Moving humanoid robots from labs and pilot projects to a scale of millions is what Brett Adcock sees as the core challenge of the next stage. This involves three closely intertwined aspects: large-scale manufacturing, ongoing learning, and social acceptance and integration.

Manufacturing Bottlenecks: Adcock admits that, unlike software that can be infinitely replicated, hardware production has physical limits. Figure has launched an internal manufacturing plant named "Baku," specifically for producing the next generation of Figure 03 robots. This robot aims to achieve "high-rate manufacturing," with costs reduced by 90% compared to its predecessors. His goal is to produce 100,000 robots in the next four years, believing this is a crucial step toward scaling to one million. "The question is, can we achieve a scale that exceeds the annual production of smartphones? This is a very difficult problem to solve, but I believe it is possible."

Learning Flywheel: The core value of large-scale deployment lies in forming a "learning flywheel." Adcock foresees that as a large number of robots are deployed around the world, they will continuously learn through interactions with the environment, with data feeding back to a central model for pre-training, thereby making the entire fleet smarter. This dynamic of "getting smarter and cheaper the more you use it" constitutes the ultimate moat in his view. He likened it: "Aside from systems like AI big models, we have never seen a technology that can interact with the world and become smarter and cheaper over time."

Safety, Privacy, and Interaction Design: As robots enter homes, safety and privacy become paramount issues. Adcock elaborated on its complexity: in terms of physical safety, robots need 360-degree perception to avoid colliding with people or pets; in "semantic safety," they must prevent incidents like knocking over candles and causing fires. To this end, Figure has established a dedicated privacy and cybersecurity team. He specifically pointed out that robots made in China may struggle to enter the European and American family markets, as establishing brand trust is critical, "this is more complex than computers and mobile phones."

Regarding the appearance and interaction of robots, Adcock has a distinctive viewpoint. He opposes giving powerful robots "cartoonish big eyes" to seem harmless, believing this is "very foolish." Robots should exhibit a sophisticated design that matches their high capabilities. In terms of interaction, he firmly believes that natural language (voice and text) will be the default interface. Controlling robots through smartphones or computers is "high-latency, low-bandwidth, and very strange," and future people will talk directly to robots or send commands via text messages. Each robot from Figure is equipped with a built-in eSIM card, having its own independent phone number.

Adcock predicts that people will soon get used to coexisting with robots. "In the future, there will be a day when you step out to run errands and see as many humanoid robots as people. It feels like pulling the future 50 years into today." He believes social acceptance will naturally increase as robots demonstrate their practicality and reliability.

Humanoid Robots: The Physical Embodiment of AGI and Future Outlook

At the climax of the dialogue, Brett Adcock elevated the significance of humanoid robots to a new height: they are the ultimate deployment carrier of AGI (Artificial General Intelligence) in the physical world.

He laid out a key argument: we are approaching digital superintelligence, but this intelligence currently "lives in servers and boxes." If there is only digital superintelligence without solving the embodiment issue, then these digital agents will require humans to perform tasks in the physical world, and humans may become "slaves of digital superintelligence." Humanoid robots, capable of performing most work of humans, are the perfect carriers to resolve this contradiction. "You can't stuff it into a vacuum cleaner at home... no other hardware can do that."

Regarding the future societal landscape, Adcock painted a transformative picture: humanoid robots will be able to accomplish more human work year by year and will ultimately surpass humans in most tasks. At that point, both household chores and professional labor will become optional for humans. Robots will significantly enhance global GDP (because they are essentially "synthetic humans") and may drastically drive down the prices of goods and services, as the costs would only include the robots themselves, energy, and possibly venue costs.

He even envisions that personal robots could be "lent out" or "rented" to generate income, just like today's Teslas are used in Robotaxi networks. Everyone will have a robot partner with "personality," capable of adjusting its sense of humor or seriousness, communicating through natural language, and continually working for you.

Finally, when asked about his expectations for 2025, Adcock's vision extends beyond the company itself. He hopes that humanity will "ascend to a higher level" in pushing technology boundaries, challenging grand ideas like underwater cities, space elevators, and orbital rings ("Neal structures") that can genuinely advance human civilization before 2050. "We need to set new benchmarks for humanity... these have become possible in our lifetime." For him, humanoid robots are not just business; they are a key to unlocking this exciting future.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。