In order to promote international exchanges and cooperation, strengthen industry-university-research collaboration, create a world artificial intelligence academic highland, and shape a central hub that connects the world with artificial intelligence industry and academic resources, the 2021 Beijing Zhiyuan Conference held by Beijing Zhiyuan Artificial Intelligence Research Institute will be held at 6 Grand opening on the 1st. Jin Wei, Deputy Mayor of Beijing, and Xu Jie, Director of the Strategic Planning Department of the Ministry of Science and Technology attended the opening ceremony and delivered speeches.
As an international high-end academic exchange event for artificial intelligence, the Beijing Zhiyuan Conference is positioned as a "top event for AI experts", and more than 200 artificial intelligence leaders including Turing Award winners such as Yoshua Bengio and David Patterson have been invited to share true expert recognition Major technical achievements and insights. As a representative of artificial intelligence companies, the fourth paradigm showcases the latest achievements and thinking of AI technology, and works with Zhiyuan Research Institute to jointly promote the application and development of AI technology.
Sign a contract with Zhiyuan Research Institute to seek AI application development
At the opening ceremony of the conference, the fourth paradigm and Xinhua News Agency, Meituan, Xiaomi, Kuaishou and other well-known enterprises jointly acted as "Enlightenment" model ecological strategic cooperative enterprises, and held a signing ceremony with Zhiyuan Research Institute to jointly develop a large-scale "Enlightenment" Intelligent application, to promote industrial agglomeration with model development and application.
The fourth paradigm will work with Zhiyuan Research Institute to jointly complete the development of large-scale pre-training models, and develop an AI product and service platform oriented to enterprise needs to help build an artificial intelligence strategic infrastructure. In addition, enterprise-level AI products and capabilities based on the fourth paradigm will provide "Enlightenment" with an AI underlying platform, and provide a full range of services and applications such as model training, testing, and deployment.
"Enlightenment" is an ultra-large-scale intelligent model developed by Zhiyuan Research Institute. It aims to create cognitive intelligence driven by data and knowledge. Through a number of innovative breakthroughs, it has formed an independent ultra-large-scale intelligent model technology innovation system with large-scale , High precision and high efficiency. At present, the parameter scale has reached 1.75 trillion, which is 10 times that of GPT-3, breaking the previous record of 1.6 trillion parameters created by foreign pre-training models, and is currently China's first and the world's largest trillion-level model.
Analyze the technological development in the AI era to promote AI application innovation
Zheng Zhao, Vice President of the Fourth Paradigm, was invited to attend the "Intelligent Architecture and Chip Forum" and published a theme entitled "From "Post-Analysis" to "Real-time Decision-making", Fourth Paradigm Machine Learning Database Design Concept and Practice" The speech introduced in detail the challenges faced by database technology in the AI era, and the evolution trend of technology.
Machine learning has been widely used in business scenarios such as advertising, thousands of people information recommendation, and real-time pricing. In these scenarios, compared to traditional solutions in which humans perform data analysis and guide business decisions, machine learning technology allows machines to automatically process massive amounts of information and provide real-time automatic decision-making capabilities. This advantage is not only reflected in relying on machines to perform large-scale The data is designed with hundreds of millions of massive strategies, which is also reflected in the machine's ability to perform faster strategy calculations and iterations.
Zheng Zhao believes that on the data supply side, compared with report analysis data, machine learning has different standards and requirements for data. Ensure the correct association of machine learning behavior feedback data, that is, closed-loop data, ensure the online and offline consistency of machine learning data calculations (Consistency), and ensure the correct timing of real-time decision data (Chronology) is to provide correct and effective data Necessary conditions for supply . For this reason, the fourth paradigm has developed and designed the core database system for machine learning applications, which has the characteristics of efficient calculation, read-write separation, high concurrency, and high-performance query, which maximizes the efficiency and performance of machine learning feature engineering. While bringing 10 times performance improvement, the cycle of machine learning application production landing was shortened from six months to two weeks.