On December 26, the 2023 Lenovo Tianxi AI Ecosystem Developer Conference with the theme of "AI Ecology and Win-win Future" was held in Beijing. Chen Xuegui, general manager of the Internet Service Division of Lenovo's consumer business group in China, said in his keynote speech that the Tianxi AI ecosystem architecture launched by Lenovo has three characteristics: comprehensive openness, device-cloud hybrid, and four-terminal integration.
Chen Xuegui, General Manager of the Internet Service Division of Lenovo China's Consumer Business Group).
Full openness is the biggest feature of Lenovo Tianxi AI ecosystem. Tianxi Ecosystem carries open cooperation with Lenovo's intelligent twins applet platform. In terms of ecological construction, Lenovo welcomes and is fully compatible with all agent applets based on any cloud large model to access the Tianxi AI ecosystem. Lenovo will adhere to the concept of comprehensive openness, focus on the basic experience of terminals, work with partners to build an AI-native application matrix, and gradually upgrade existing traditional applications with AI.
In terms of device-cloud hybrid. Although the traditional cloud model has the advantages of high performance and generalization capabilities, with the gradual deepening of the application field, the disadvantages of high deployment and inference costs, high energy consumption, hidden dangers of privacy leakage, and inability to provide personalized services also hinder its further large-scale application. The advantages of high real-time, low energy consumption, high security, and low cost of the device-side personal private model can make up for the shortcomings of the cloud large model. The device-cloud hybrid model can complement each other's strengths, learn from each other's strengths, and match high and low, which will greatly reduce the cost of using AI and truly realize the inclusiveness of AI.
In terms of four ends in one. The cross-terminal design of Tianxi AI architecture includes AIPC, AI tablet, AI mobile phone and AIoT terminals, and each end has the capabilities of intelligent perception, computing power scheduling, and inference acceleration. The personal agent will play the role of the user's personal AI assistant and shuttle through the four types of AI terminals to achieve cross-terminal relay and integrate them. The personal agent will break the limitations of the device-side system, communicate with the user in multi-modal natural language interaction, understand the user's intention, realize the orderly distribution of tasks, and feedback the final result to the user.