Among many economic indicators, only one is a true risk-free rate, which is the growth rate of scientific and technological productivity. It reflects the core role of technological innovation in improving production efficiency and promoting economic development. It is not affected by individual credit risks and is a direct reflection of the technological progress and production potential growth of the entire economy.

Every technological revolution has generated a large amount of incremental value, and continuous technological innovation has brought value growth far higher than the return on capital. Thanks to Moore's Law, the capabilities of AI have increased exponentially, and productivity has been greatly improved in any link where AI can participate. A report released by McKinsey last year pointed out that "generative AI may add $2.6 trillion to $4.4 trillion to the global economy each year", which is equivalent to 15% to 40% of the economic impact of the entire AI technology.

In the wave of artificial intelligence, the development of humanoid robots has always attracted much attention. They not only represent the forefront of technological progress, but also are partners in future intelligent life. However, to make these robots act naturally like humans, a series of complex control problems need to be solved.

Hierarchical world model, data-driven humanoid robot control

Recently, LeCun proposed a new data-driven humanoid robot control solution. The core of this solution is the hierarchical world model, which achieves more natural and efficient motion control by training high-level and low-level intelligent agents, and achieves efficient completion of complex tasks through the decision-making of high-level intelligent agents and the execution of low-level intelligent agents.

In LeCun's research, the "Puppeteer" model used a simulated 56-DoF (degrees of freedom) humanoid robot to generate high-performance control strategies in 8 tasks. These strategies not only have excellent performance, but also natural movements, showing the ability to traverse challenging terrain. LeCun's layered world model and the Puppeteer project provide new ideas and methods for the control of humanoid robots.

On this basis, WorldBrain will further explore and realize the potential of humanoid robots and promote the development of artificial intelligence technology. The uniqueness of WorldBrain's hierarchical world model is that it combines the hierarchical and distributed processing capabilities of neural networks to process various types of information and complex relationships. The model emphasizes large-scale data processing and can extract information from multiple sources, laying the foundation for more comprehensive understanding and decision-making. By combining advanced artificial intelligence technology, neuroscience and blockchain technology, WorldBrain is expected to build a comprehensive and decentralized artificial intelligence system.

Modeling large-scale intelligent neural network systems by imitating the human brain

At present, the essential logic of the development of artificial intelligence at this stage is that the machine summarizes the rules from a large amount of "specific" data, summarizes certain "specific knowledge", and then applies this "knowledge" to real-life scenarios to solve practical problems. Although AI cannot completely replicate human intelligence, it simulates certain cognitive functions of humans in specific tasks, such as perception, reasoning, and problem solving.

Humans have developed a mental model of the world based on what they can perceive with their senses. This model includes the abstraction and generalization of the structure, attributes, behavior, and laws of the world. The core of the WorldBrain project is to build a distributed intelligent neural network system that can simulate the human brain. It uses "WorldModel" and combines artificial intelligence, neuroscience, and blockchain technology to create a comprehensive operating system for future AGI (artificial general intelligence) terminals.

The TD-MPC2 algorithm played a key role in LeCun's research. Through two stages of training, it enables two agents to work together to generate natural and efficient actions. The successful application of this algorithm provides valuable experience for the WorldBrain project. WorldBrain can learn from this algorithm to further optimize its neural network system and make it more adaptable to complex and changing environments and tasks.

Intelligence that can understand, reason, and interact with the physical world

AI is a broad field that includes many technologies and applications such as machine learning, deep learning, and natural language processing. Robots need a deeper understanding of language and the physical world, and reasoning about the physical world requires a detailed understanding of the objects in the environment and how these objects respond to the robot's actions.

WorldBrain establishes the initial world (reference frame) for "WorldModel", thus realizing WorldBrain's ability to model the world like the human brain, and giving robots the ability to visually recognize, think, act and learn, allowing robots to work effectively in various scenarios. The WorldBrain simulation dimensions are as follows:

Perception and behavior simulation: WorldBrain can mimic human cognitive abilities and behavioral performance through perception and behavior simulation. In terms of perception, WorldBrain uses computer vision and speech recognition technology to perceive visual and auditory inputs to obtain environmental information. In terms of behavior simulation, WorldBrain uses robotics or virtual simulation environments to simulate human actions and interactions and interact with the environment in real time. This enables WorldBrain to perceive and interact with the environment like a human brain embodied system.

Movement and body representation simulation: WorldBrain can mimic human movement and physical expression through movement and body representation simulation. By controlling the robot's movement and posture, WorldBrain can simulate how humans move and manipulate in space. This includes precise manipulation of the robot's arms and posture adjustment of the robot's body, allowing WorldBrain to perform complex manipulations and interactions like humans.

Spatial and contextual modeling: WorldBrain can mimic human understanding and reasoning about the environment through spatial and contextual modeling. By modeling the spatial structure and contextual information of the environment, WorldBrain can acquire knowledge about objects, space, and time. This can involve using maps, sensor data, and contextual information to build an internal representation of the environment to better understand and reason about tasks related to the environment.

The distributed brain is another unusual embodiment: the human neocortex has about 150,000 cortical columns, each of which models part of the world it can perceive. The cortical columns of the WorldBrain smart terminal do not have to be adjacent to each other like the cortical columns in the biological brain. A smart terminal may have millions of cortical columns and thousands of sensor arrays, or it may only have one simulated neuron and a few sensor arrays. The DePIN network uses blockchain technology to coordinate and adjust millions of individuals to deploy and run trustless, permissionless, and programmatic infrastructure, laying a solid foundation for WorldBrain's hardware deployment.

The goal of WorldBrain is not only to achieve technological breakthroughs, but also to make artificial intelligence truly serve humans and improve people's quality of life. Only when robots' movements are more natural and more in line with human expectations and habits can they better integrate into human society and become assistants and partners in people's lives.

In the future, we have reason to believe that WorldBrain will not only be a technical project, but will also become a bridge connecting humans and intelligent machines, ushering in a new era of greater intelligence and interconnection. With the continuous advancement and innovation of technology, WorldBrain and humanoid robots will jointly usher in a future full of infinite possibilities.