What is zero one thousand things?
01Wanwu is an AI2.0 incubation project of China Innovation Works, led by Dr. Li Kaifu himself, and is committed to building China's leading large-scale model technology team.
Large-scale models refer to artificial intelligence models that can process massive amounts of data. They have strong versatility and creativity, and can perform various tasks across domains, media, and languages. They are the core technology of the AI2.0 era.
01Wanwu is taking the road of independent research and development of large models. It has completed internal testing at the level of tens of billions of parameters, and plans to increase the scale of the model to 30 billion to 70 billion parameters in the future. In addition, the company plans to launch an open-source version of the model to share technical results with the community. The vision of 01Wanwu is not only to be a general-purpose large model, but also to be a multi-modal model that can process multiple data types such as pictures, videos, and 3D.
What are the differences between AI2.0 and AI1.0?
- AI1.0 refers to the computer vision technology based on the CNN convolutional neural network model as the core. It mainly realizes the perceptual intelligence of AI. Machines can surpass humans in the fields of computer vision and natural language understanding, but it requires a large amount of labeled data and professional knowledge. Programmers, limited application scenarios.
- AI2.0 refers to the large-scale model technology based on super massive data training, which can realize cross-domain, cross-media, and cross-language general artificial intelligence, with stronger understanding, reasoning and creativity, without the need for manual labeling of data and professional programmers. It can automatically adapt and execute various tasks, and has a wide range of application scenarios.
- AI1.0 is like inventing electricity, and AI2.0 is the power grid. AI1.0 provided the foundation and inspiration for AI2.0, but AI2.0 has achieved a qualitative leap and opened a new era.
What are the advantages of 01Wanwu?
01Wanwu has the following advantages:
- Team Strength: The company is headquartered in Beijing, but it is not restricted by region, and looks for the best talents from all over the world. The team has dozens of professionals who are good at large model technology, artificial intelligence algorithms, natural language processing and other fields. The core team members of 01Wanwu are industry elites, including former vice presidents of Alibaba and Baidu.
- Technical strength: 01Wanwu has the ability and experience to independently develop large-scale models, and has achieved breakthroughs at the level of tens of billions of parameters, and has higher goals and plans. Zero One Wanwu also has the technical advantages of multi-modal models, which can handle multiple data types and realize richer and smarter application scenarios.
- Data strength: 01Wanwu makes full use of the advantages of Chinese language data. Chinese is the most spoken language in the world, and it is also one of the most complex and richest languages. Chinese data is not only huge in quantity, but also high in quality, which can provide better training materials and broader application space for large models.
What are the challenges of 010,000 things?
Zero One Wanwu is also facing challenges in the following aspects:
- Hardware cost: Large models need to consume a lot of computing resources and storage space, which means that huge hardware costs and operation and maintenance costs need to be invested. At present, there are problems of insufficient hardware supply and rising prices at home and abroad, which bring difficulties to the development and deployment of large models.
- Technical Difficulty: Large models not only need to train massive amounts of data, but also need to optimize algorithms and architectures to improve efficiency and performance. At the same time, large models also need to solve issues such as interpretability, reliability, and controllability to ensure their safety and compliance.
- Application landing: Although the large model has strong versatility and creativity, it needs to be adapted and optimized to meet the needs and expectations of users before it can be applied to actual scenarios. At the same time, large models also need to communicate and collaborate with the industry and society to establish a good ecology and influence.
What are the prospects of Zero One and Everything?
01Wanwu has a grand vision and lofty goals. It hopes to promote the development and application of AI2.0 in China through large-scale model technology, and bring more value and possibilities to human society.
Dr. Kai-Fu Lee believes that self-developed large models are a necessary condition for the realization of AI2.0. We cannot completely rely on open source models and small models, and we must establish our own technical advantages and ecological barriers. He also said that 01WOOW will be committed to creating an open, shared, and collaborative large-scale model platform, and will explore and create the future of AI2.0 together with researchers, developers, and entrepreneurs around the world.
China's current situation and prospects in the field of AI2.0
AI2.0 refers to the large-scale model technology based on super massive data training, which can realize cross-domain, cross-media, and cross-language general artificial intelligence, with stronger understanding, reasoning and creativity, without the need for manual labeling of data and professional programmers. It can automatically adapt and execute various tasks, and has a wide range of application scenarios.
AI2.0 is a new era, and both China and the world are facing opportunities and challenges. Based on some information on the Internet, I have summarized the following points:
One of China's advantages in the field of AI2.0 is that it has massive data, which is an important basis for training large models. China's digitalization process and Internet applications are very extensive, involving various industries and fields, providing a rich data source for AI2.0.
The second advantage of China in the field of AI2.0 is that it has strong policy support and social needs. The Chinese government attaches great importance to the development of AI, and has formulated the "New Generation Artificial Intelligence Development Plan", which puts forward a series of goals and measures, and provides a good environment for AI2.0. At the same time, Chinese society has a strong demand for AI applications, such as smart cities, smart medical care, and smart manufacturing, providing a broad market for AI2.0.
One of China's shortcomings in the field of AI2.0 is the lack of core technology and innovation capabilities. Although China has made some progress in the research and development of large models, it still lags behind the international leading level, and many core technologies and algorithms still rely on foreign open source or commercial platforms. At the same time, China's innovation capabilities in the field of AI2.0 are not strong enough, lacking original and disruptive breakthroughs.
The second shortcoming of China in the field of AI2.0 is that it faces the challenges of international competition and cooperation. AI2.0 is a global trend, and all countries are accelerating deployment and investment, forming a fierce competition situation¹³. At the same time, AI2.0 also requires cooperation and exchanges on a global scale, sharing data, resources and experience, and solving common problems and challenges¹³. How to find a balance between competition and cooperation is a problem that needs to be considered and dealt with.
China's successful cases in the field of AI2.0
Simou Technology, a unicorn company focusing on industrial vision AI, was founded at the end of 2019. It has entered more than 1,000 sub-industries and served more than 100 world-renowned companies and institutions such as Carl Zeiss, Airbus, and Bosch. Simou Technology has created the world's first industrial AI operating system SMore AIOS, formed the SMore Factory intelligent manufacturing solution, established a complete software and hardware ecosystem, and has full-stack industrial AI software and smart device development capabilities, as well as production line-level deployment capabilities. The goal of Smart Technology is to break away from the data limitations of a single algorithm field and create an AI2.0 era with a systematic architecture.
Founded in 1999, iFLYTEK is a world-leading provider of intelligent voice and artificial intelligence services. It has won the championship of international voice competitions for 12 consecutive years. HKUST Xunfei proposed the AI2.0 strategy, concentrating resources on 8 core product lines including education, consumers, open platforms, smart cities, and medical care. In the ever-expanding application scale, "algorithm + data" iterations are carried out. Constantly evolving and landing. HKUST Xunfei's AI2.0 emphasizes quality and efficiency improvement, per capita efficiency improvement, and the implementation of technology to generate actual benefits.
Academician Pan Yunhe's team is a research team led by Pan Yunhe, academician of the Chinese Academy of Engineering and professor of Zhejiang University, with profound accumulation in the fields of digital library, big data intelligence, swarm intelligence, and cross-media intelligence. Academician Pan Yunhe's team proposed the concept of transforming numbers into knowledge, actively established a knowledge center, and realized cross-database retrieval and knowledge services for heterogeneous data in different fields. Academician Pan Yunhe's team believes that entering the stage of AI2.0, artificial intelligence has been upgraded, and the core feature of the 2.0 era is the integration of human-machine intelligence.