The world of AI is advancing at an astonishing pace, creating a new intelligent era.
The topic of artificial intelligence during the 2024 Spring Festival has been going viral. The US company OpenAI officially released its first text-video generation model, Sora, which can create realistic and imaginative scenes based on text instructions. Thanks to OpenAI, Microsoft's market value has exceeded 3.1 trillion US dollars, surpassing Apple to become the world's most valuable company.
Subsequently, the US technology giant Google announced the launch of Gemma, the world's most powerful and lightweight open source model series, with performance far exceeding the mainstream open source models Llama and Mistral, and directly topped the Hugging Face open source large model rankings.
In terms of artificial intelligence BCI, Musk announced that the first human to be implanted with a Neuralink brain chip has fully recovered. The patient can control the mouse with just his thoughts, ushering in a new era of brain-computer interfaces.
NVIDIA released its financial report for the fourth quarter of 2023, which showed that NVIDIA’s revenue in the fourth quarter of last year was US$22.1 billion, exceeding market expectations. Catalyzed by technological innovation, global artificial intelligence concept stocks have surged recently, and many stocks have attracted attention from institutions. Goldman Sachs predicts that global artificial intelligence investment will grow to US$158.4 billion in 2025, which is just the tip of the iceberg. According to the latest artificial intelligence statistics, as of 2023, the global artificial intelligence market size is estimated to be US$454.12 billion. At an annual growth rate of 19%, the artificial intelligence market may be worth more than $2.5 trillion by 2032.
We can be sure that this society will eventually move towards artificial intelligence, so where will artificial intelligence go next?
Artificial intelligence and accelerated computing are jointly transforming the technology industry. The next wave of artificial intelligence will be a new type of artificial intelligence called embodied AI, that is, intelligent systems that can understand, reason and interact with the physical world. Nvidia founder and CEO Jensen Huang pointed out the direction.
In exploring the integration of intelligence and reality, WorldBrain insists on long-term research, builds technical reserves, and achieves breakthroughs and new developments. WorldBrain is currently the only outstanding innovative project that integrates the concept of embodied intelligence into Crypto. Through a large-scale intelligent neural network system that imitates human brain modeling, WorldBrain introduces intelligence into specific objects and machines, giving them the ability to perceive, understand and interact. Its uniqueness lies in leading intelligence from the virtual world to reality and exploring the integration of intelligence and reality.
The Origin of Embodied Intelligence
The idea of embodied intelligence originated at the beginning of artificial intelligence.
The concept of embodied intelligence was first proposed by Alan Mathison Turing, the father of artificial intelligence, in 1948. Embodied means human body, and intelligence means intelligence. When combined together, they form embodied intelligence that is inseparable from the human body. At the same time, Turing also proposed another opposite concept, Disembodied Intelligence, which is intelligence outside the body and away from the human body.
In 1950, Turing proposed two possible development paths for artificial intelligence in his groundbreaking paper "Computing Machinery and Intelligence": one is to focus on the intelligence required for abstract calculations (such as playing chess), and the other is to equip machines with the best sensors, giving them the ability to communicate with humans and learn like babies. These two paths gradually evolved into disembodied intelligence and embodied intelligence.
What is embodied intelligence?
Embodied Artificial Intelligence (EAI), also known as "embodied AI", refers to an intelligent system that perceives and acts based on the physical body. It obtains information, understands problems, makes decisions and takes actions through the interaction between the intelligent agent (which can be biological or mechanical) and the environment, thereby generating intelligent behavior and adaptability.
Embodied intelligence can be briefly understood as an artificial intelligence research paradigm that emphasizes that an intelligent body with a physical body acquires intelligence by interacting with the physical environment through the creation of an intelligent body that combines software and hardware. Let's break it down: one is "embodiment" and the other is "intelligence". The basic meaning of embodiment refers to the dependence of cognition on the body, that is, the body has an impact on cognition. Cognition cannot exist independently from the body. In addition, the opposite concept of "embodiment" is "disembodiment", which refers to the decoupling of cognition from the body. The understanding of intelligence in the field of artificial intelligence is mostly biased towards human abstraction, logical ability and adaptability.
Exploring the new dimension of AI: WorldBrain’s new perspective on embodied intelligence
Compared with traditional abstract artificial intelligence, embodied intelligence is a form of intelligence that can perceive and interact with its environment and has self-perception capabilities. Embodied intelligence can exist in many forms, such as robots and autonomous driving. They can not only understand the information in the environment, but also obtain and process information from complex environments at a higher level.
WorldBrain is an unprecedented distributed intelligent neural network system based on the WorldModel world model, which simulates the operating mechanism and thinking process of the human brain to better understand the connection between artificial intelligence and cognitive science. It is an important part of the future realization of "strong artificial intelligence". By adopting complex computational models, including modeling of neural networks, information processing, learning and decision-making, it helps to deeply study the basic principles of human cognition and provide more effective inspiration for AI systems.
WorldBrain has brought humans a more intelligent and efficient way of information processing, and built a new intelligent ecosystem for the future digital society.
First, embodied intelligence can understand and respond to its own environment. This is achieved through advanced perception technologies, such as visual sensors, voice recognition systems, or tactile sensors, which enable embodied intelligence to obtain information from the environment. WorldBrain draws on the mechanisms of brain maps and reference frames to obtain more accurate spatial perception and environmental understanding capabilities through perceptual inputs (such as visual and position sensor data) by using the maps and reference frames in the human brain as the basis for learning world models. This will enable WorldBrain to better interact with the physical world and achieve more advanced cognitive and intelligent functions.
Second, embodied intelligence can improve itself and learn new skills through machine learning. This makes embodied intelligence not just a static system, but a dynamic system that continuously improves and learns through big data and deep learning techniques. WorldBrain can be quickly trained in an unsupervised manner to learn a compressed spatial and temporal representation of the environment. By using features extracted from the world model as input to the agent, a very compact and simple strategy can be trained to solve the desired task.
Finally, embodied intelligence can interact naturally with users, understand and handle complex tasks. Human-computer interaction has risen to a new level. Unlike the current popular large language model multimodal interaction between text, pictures, audio, video, etc., the important manifestation of embodied intelligence is multimodal + cross-modal perception interaction, especially cross-modal perception of vision, language and body behavior; this requires not only artificial intelligence, but also human-computer interaction, natural language processing and other technologies. WorldBrain can imitate human cognitive ability and behavioral performance through perception and behavior simulation, and imitate human motor ability and physical expression through movement and body representation simulation, so that WorldBrain can perform complex operations and interactions like humans.
Embodied intelligence is a possible path to producing super artificial intelligence.
Embodied intelligence is the new frontier of artificial intelligence and a key step in the evolution towards higher levels of intelligence such as perception, understanding, learning, and adaptation.
The "Zhiyuan Artificial Intelligence Frontier Report" (2021-2022) points out that in the next three years, embodied models based on virtual world and real-time space-time environment training will achieve great development, such as autonomous driving, robots, digital people in games, etc. In the next five to ten years, ultra-large-scale pre-trained models (information models) and embodied models will be combined to become "digital supermen", which will be stronger than previous humans in terms of knowledge ability and degree of interaction with the environment. Embodied models and robots will also be combined to produce unmanned systems with stronger capabilities than humans in the physical world, namely "embodied supermen". Optimistic estimates show that in the next thirty years, digital supermen and embodied supermen may be combined to eventually give birth to super artificial intelligence.
Sora's one-shot and multi-shot images are powerful, realistic and imaginative. However, Sora does not understand the order of glass breaking and liquid spilling, does not understand the common sense that ants have six legs, and cannot reason about time and causality. Trying to understand the world through fitting methods, pure fitting routes cannot become AGI. AGI ultimately needs to touch and learn the world through self-supervision.
The ultimate dream of artificial intelligence is to achieve super artificial intelligence.
Embodied intelligence is a possible path to produce super artificial intelligence. Embodied intelligence is an important way to improve the cognitive ability of current "weak artificial intelligence". Artificial intelligence can learn and improve from real physical or virtual digital space through channels of interaction with the environment.
WorldBrain will continue to advance along the path of Embodied Intelligence pioneered by Turing, pursuing a closer and more natural connection between artificial intelligence technology and the real world. Although it is still in its early stages, its exploration of breaking the barriers between artificial intelligence and the real world has shown unlimited potential.
