Behind AIGC, large models, and large computing power are extremely popular, what are their application scenarios, and what rules can be followed in business practice? Multi-cloud practice, business architecture evolution, R&D efficiency and other technical hotspots that seem to be "clichés", are there any new development trends in the current era of advocating business innovation?
On June 16-17, 51CTO gathered 50+ lecturers, experts with forward-looking vision and practical experience from AI, cloud computing, big data, architecture and other fields, as well as 800+ participants, to hold the WOT Global Technology Innovation Conference in Beijing . Based on the principles of real scenarios, real battles and real methods, we will explore together how new technologies can help enterprises achieve a leap and improvement in digital capabilities.
At the beginning of the conference, Xiong Ping, founder, chairman and CEO of 51CTO, said in his opening speech: It is the mission of 51CTO to help 100 million digital talents grow for life. We are full of confidence in digital talent training. Statistics show that the annual increase of China's digital talents reaches one million, and the stock exceeds 20 million. Learning is becoming the main way to accelerate the digitalization process in this world.
Since its establishment 18 years ago, 51CTO has always adhered to the concept of "Technology Achieves Dreams", maintained keen technical insight, explored IT technology development trends, focused on the implementation of cutting-edge technology application scenarios, provided technical personnel with high-quality content and services, and empowered global technology People grow. So far, 51CTO has 20 million registered users and more than 12,000 technical experts.
How does the large model empower thousands of industries?
Large-scale models are a very hot topic in the current technology circle, and the popularity has lasted for several months without any signs of decay. Zhou Ming, founder and CEO of Lanzhou Technology, vice chairman of CCF of China Computer Federation, and chief scientist of Sinovation Ventures, delivered a keynote speech entitled "How Large Models Can Empower Thousands of Industries".
In Zhou Ming's view, the large model has brought about a leapfrog development of cognitive intelligence technology. From AI large model 1.0 to 2.0, from simple capabilities + dedicated models for specific tasks, to complex capabilities + general-purpose models for general tasks, large models have promoted language understanding, multi-round dialogue, and problem solving into the era of basic availability. In addition, the development of large models can effectively solve the problem of fragmentation of NLP tasks and greatly improve the efficiency of research and development, which also marks that NLP has entered the stage of industrialization and implementation.
Driven by the large model, the general large model has moved towards the industry large model, which has accelerated the process of digital transformation of enterprises.
In Zhou Ming's view, the large model will become the basic model of all walks of life, and become the center of enterprise services through cloud computing, local deployment and terminal. Taking the financial industry as an example, he introduced how the "Mencius Financial Big Model" can help financial customers achieve business innovation through the big model.
How does large computing power help AIGC?
The continuous popularity of large-scale models has led many domestic technology companies to devote themselves to the construction of large-scale models. They have excellent technical teams, as well as rich industry knowledge and innovative ideas, but the bigger challenge is that they often lack systems that have been practiced in large-scale scenarios engineering capabilities.
In his sharing, Luo Hao, head of cloud infrastructure products of Volcano Engine, introduced how Volcano Engine adapts to the requirements of the era of large models and large computing power during the evolution of infrastructure. Volcano Engine continues to provide high stability and cost-effective AI infrastructure for large model customers . Its self-developed DPU network performance reaches 50 million pps forwarding capability, and the delay is as low as 20us. The performance of various computing instances based on the self-developed DPU has also been significantly improved. For example, the performance of GPU instances suitable for large-scale distributed parallel training scenarios has increased by up to 3 times compared with the previous generation instance cluster.
Why do we need "big models"?
Chen Yuheng, the co-founder of SenseTime, shared his thinking and product practice in the era of large models. According to him, in the past four years, the average annual compound growth rate of large model parameters has exceeded 400%, and the demand for computing power has increased by more than 150,000 times, far exceeding Moore's Law.
Why do we need "big models"? Because “big” equals better generalization — generalization refers to the ability of a machine learning model to adapt and predict accurately when faced with new data.
SenseCore has a computing power of 5000 Petaflops, supports simultaneous training of 20 ultra-large models with hundreds of billions of parameters, and provides customers and ecological partners with a full-chain MaaS large model as a service. As of May this year, SenseTime has served more than 40 core customers, including more than 10 large-scale model customers, and has delivered large-scale models in more than 20 landing scenarios.
How to design a large-scale payment platform architecture?
Compared with other Internet application platforms, the payment platform requires higher efficiency, usability, security and scalability. In the new era of cloud native and AI, how to build and optimize an excellent Internet payment platform by using new technologies and ideas is a question that CTOs and architects in the payment industry think about day and night.
In his sharing, Chen Bin, CTO of NETSTARS, explained in detail how to do a good job in the architecture design of large payment platforms.
About the conference
In addition to sharing at the main venue, this conference also set up 12 parallel sub-venue activities. For AIGC, large-scale models, and large computing power, which are the most popular technologies at present, all practitioners who have studied and cultivated in these fields are invited to give calm analysis and forward-looking predictions.
For topics such as multi-cloud practice, business architecture evolution, and R&D efficiency improvement, which have long attracted the attention of technical people, the branch venue also focused on practice and delivered high-quality content.
Since 2012, 51CTO has contributed 19 WOT Global Technology Innovation Conferences for 11 consecutive years. In the past 11 years, some technologies have risen like myths and then disappeared like bubbles; some technologies have become more and more valuable after being tempered by time; some technologies are full of controversy and the controversy continues.
In 2023, which is full of challenges and full of vitality, we will use this technology conference to witness the power and value of technology.