According to the report from Zhijun Finance and Economics App, Figure AI, a startup company of AI humanoid robots, which is supported by many technological leaders such as OpenAI, Microsoft, NVIDIA and the founder of Amazon, Jeff Bezos, released the second generation humanoid robot Figure 02 at eastern time on Tuesday. As the successor of Figure 01, which was launched in 2023, this is considered the most advanced humanoid robot currently available. This new robot is fully integrated with the multi-modal AI model launched by OpenAI. It is widely regarded as a “walking ChatGPT” with many powerful capabilities, such as mimicking human behavior, deep learning and thinking, and communicating with humans naturally and efficiently.

According to Figure AI, this is the most powerful and comprehensive humanoid robot launched by Figure AI to the global market because it combines the flexibility of human body shape with the advanced AI model of OpenAI. The powerful imitation and deep learning capabilities of this robot mean that it can accurately perform various complex and dangerous tasks in enterprise production and manufacturing activities. It plays a more intelligent role in assisting humans to improve productivity in human production activities. In the near future, it is expected to be widely used in households and become a consumer electronics product comparable to iPhone and iPad.

Some technology industry researchers commented that the release of Figure AI's new second-generation humanoid robot Figure 02 may mean that human society is gradually entering the “smart machine era”. The “smart machine era” usually refers to an era in which AI humanoid robots and artificial intelligence technology are widely used in human society.

Many science fiction and film works have predicted that the “smart machine era” may come soon. In this era, AI humanoid robots will be able to make autonomous decisions, learn and adapt to complex environments and communicate with humans naturally and efficiently, accurately completing many production and manufacturing activities. This technological progress may have a positive and far-reaching impact on economic growth and human social productivity.

The timetable for the wide release of Figure 02 has not been announced yet, but Figure AI wrote in the latest introduction: "Figure robot perfectly combines the flexibility of human form with the most advanced artificial intelligence technology, and will perform various tasks in business applications and households soon. "

Figure 02, which integrates OpenAI's large model, can be regarded as the "most powerful edge AI".

The most eye-catching feature of Figure 02's update function is undoubtedly the full integration of OpenAI's multi-modal large model, which was achieved through a long-term cooperative relationship with OpenAI. In February, OpenAI helped Figure AI to raise about $0.675 billion in Series B financing and valued this Silicon Valley startup at $2.6 billion.

The advent of AI large models has historic significance for the entire robot industry, and humanoid robot developers are particularly interested in this technology. One of the main selling points of this appearance is that it can communicate and work efficiently with human colleagues in factory workshops-provided that appropriate safety measures are taken. Figure 02 has a deep learning and imitation ability based on OpenAI's large model, and is equipped with speakers and microphones, which can talk normally with people at work.

Multi-modal large models like OpenAI GPT-4o and Google Gemini are highly respected for their incredibly powerful natural language abilities and efficient problem-solving capabilities, creating a new chapter in the field of intelligent assistants and chatbots. It is an obvious trend to equip these functions for these humanoid robot systems: it helps humans guide robots conveniently. With just one sentence or one action, robots can react instantly like humans, and it also increases transparency for what robots do at any given time.

When the AI chip deployment scale for handling massive parallel computing has reached the basic power demand and basic performance support, according to the trend of technological development in recent years, AI large models will eventually be integrated into these terminals, including consumer electronics such as smartphones and humanoid robots, as well as applications such as software systems for electric vehicles and industrial production. Figure AI's release of the second generation humanoid robot Figure 02 can be regarded as the "most powerful edge AI" seen so far.

Compared with cloud-based AI, edge AI with significant advantages such as high efficiency, fast response, and personalization is more in line with consumers' actual needs, which will inevitably bring about a surge in demand for inference chips. In comparison to AI training, the AI inference field has far less demand for GPU parallel computing power under the application background of "massive data bombardment". The inference process involves using a trained model to make decisions or identifications. CPUs are the core of the processor, and they are extremely good at handling complex logic processing tasks and control flow tasks.

In the work of integrating AI big models into humanoid robots, Figure is certainly not alone. At the end of last year, the robot company Agility showcased the work it has been doing, which is to comprehensively improve the efficiency of communication between humans and robots using generative artificial intelligence. The use of neural networks was a key project for the Google Everyday Robots team prior to its closure. At the same time, Tesla CEO Musk's Grok AI and Optimus ("Transformers" humanoid robot) are projects that will undoubtedly be connected sooner or later.

Before investing in Figure AI, OpenAI was very active in the field of humanoid robots. The company had invested in Norwegian humanoid robot company 1X. But without a doubt, Figure AI, with its impressive robot appearance, flexible walking pace, and powerful imitation and learning ability, has been the most active player in the industry over the past year. Other top technology companies investing in Figure AI include Microsoft, Amazon, Nvidia, and Intel, among others.

Figure AI has recently begun collaborating with BMW on a pilot program for car manufacturing. In June, the company released a video showing early robots performing tasks autonomously on the floor with the help of the neural network system. The company noted that the Figure 02 robot had visited an auto manufacturer factory in Spartanburg, South Carolina, for imitation, training, and data collection.

Collaboration between humanoid robots and auto manufacturers has become increasingly frequent recently. Agility, Apptronik, and Sanctuary AI have announced similar pilot programs with car manufacturers. Tesla CEO Musk has long viewed humanoid robots like "Transformers" as a key factor in improving Tesla's automotive production capacity, while Hyundai, the owner of Boston Dynamics, is eyeing Boston Dynamics' proprietary humanoid robots.

The ability to communicate with humans is an important part of the "bottom-up hardware and software redesign" referred to by Figure between 01 and 02. The list also includes six RGB cameras, a built-in visual language model, and improved CPU/GPU hardware and robot arms that are becoming more and more flexible, approaching human-like flexibility.

"AI Godfather" Huang Renxun: The next wave of AI will be in the field of robotics

With the emergence of OpenAI's Sora, a text-to-living video AI big model that can understand and simulate the physical world of motion, AI big models with more powerful physical world modeling capabilities and more comprehensive reasoning abilities based on multi-modal standards may promote the prosperity and development of the humanoid robot industry.

Huang Renxun, the founder and CEO of Nvidia, which has the title of "AI Godfather," recently said: "The next wave of AI will focus on the field of robotics, with one of the most exciting developments being humanoid robots.""We are advancing the entire NVIDIA robot technology stack, opening up technology platforms for global humanoid robot developers and companies to use the platforms, accelerated libraries, and AI big models that best suit their needs."

In recent years, Nvidia has undoubtedly focused its R&D efforts and expenditures on the field of humanoid robots, believing that humanoid robots will be the core application scenario of AI technology. At the end of July, Nvidia announced a set of services, models, and robot computing platforms for leading developers, AI big model developers, and software manufacturers in the humanoid robot industry to develop, train, and build the next generation of humanoid robots.

Nvidia's new services and products include the new NVIDIA NIM microservices and frameworks for humanoid robot simulation and deep learning, the NVIDIA OSMO orchestration service for running massive multi-stage humanoid robot workloads, and the remote operation workflow that supports new AI technologies and humanoid robot simulation, allowing developers to train humanoid robots with a small amount of human demonstration data.

It is understood that Nvidia's newly launched NIM microservices provide pre-built containers supported by powerful inference computing software from NVIDIA, enabling humanoid robot developers to reduce the deployment time of robot simulation solutions from up to several weeks to just a few minutes. The two new NVIDIA AI microservices will allow robot development experts to enhance the simulation workflow of generative physics AI in NVIDIA Isaac Sim, a robot simulation project reference application built on the NVIDIA Omniverse platform.

A recent forecast report from Markets And Markets shows that the global humanoid robot market is expected to be only about $1.8 billion by 2023, but is expected to rapidly expand to $13.8 billion by 2028, with a compound annual growth rate of over 50%. In June, Musk made a stunning statement at Tesla's shareholder meeting, saying that the "Transformers" humanoid robot could be the core catalyst for Tesla's market cap to reach $25 trillion.

In July, Musk emphasized at Tesla's earnings call that the "Transformers" humanoid robot will be put into production next year. He expects that by 2025, thousands of "Transformers" humanoid robots will perform important production and manufacturing tasks for the company, and the second version is expected to be sold to external companies in 2026. In the future, "Transformers" humanoid robots will be just as important a revenue generator for Tesla as Robotaxi.