The future shape of AI PC: Yang Yuanqing's outlook

[November 22, 2023, Beijing] On November 22, the "Finance" Annual Conference 2024: Forecast and Strategy" was held in Beijing. Yang Yuanqing, Chairman and CEO of Lenovo Group, published "Accelerating New IT Technology" at the annual conference "Innovation to Promote Inclusive Benefits of Artificial Intelligence" keynote speech. In his speech, he systematically elaborated on Lenovo's insights and business layout on the wave of artificial intelligence large-model technology, and also for the first time fully defined the five characteristics of future AI PCs (artificial intelligence computers).

(Lenovo Group Chairman and CEO Yang Yuanqing delivered a keynote speech on "Accelerating New IT Technology Innovation and Promoting Inclusive Benefits of Artificial Intelligence" at the 2024 "Finance" Annual Conference)

Two industry trends drive the emergence of artificial intelligence PCs

Yang Yuanqing mentioned in his speech that the rich and diverse large public models have brought efficiency and convenience to people, but at the same time they have also brought about very practical problems, that is, how to enjoy the efficiency dividends brought by large models while also Can effectively protect personal privacy and data security. Yang Yuanqing believes that through the mixed use of public large models and personal large models, such "both need and need" can be achieved. The future artificial intelligence large models will be private large models (personal/enterprise-level large models) and public large models. Hybrid artificial intelligence where models coexist.

He explained that the personal large model refers to the basic artificial intelligence model deployed on personal smart devices or home servers. It uses personal data stored locally for reasoning and learning. Compared with public mockups, personal mockups can not only provide answers and create content through dialogue, but are also more accurate and relevant. It can even predict tasks and proactively find solutions based on the user's mental model. Personal data will not be shared or sent to the public cloud, ensuring personal privacy and data security

At the same time, building large models requires training and inference on large amounts of data, which places extremely high demands on computing power. Yang Yuanqing said that with the further development of artificial intelligence applications, the proportion of computing load for large-scale model training and inference will also undergo major changes, and new requirements will be put forward for the allocation of computing resources. Currently, the number of users using large models is relatively small, and most large models are trained on public cloud platforms with strong computing capabilities. However, in the future, as the user scale expands, the computing resources required for inference will increase rapidly and exceed those required for training. At that time, relying solely on the public cloud to complete all training and inference tasks will lead to increasingly prominent problems of inefficiency and high costs

Therefore, Yang Yuanqing believes that whether it is for data security and privacy protection, or for higher efficiency and lower cost in response to user needs, the computing load of large models will gradually shift from the cloud to the edge and device sides. Going forward, more and more artificial intelligence reasoning tasks will be performed on the edge and device side, making personal large models more necessary and possible.

Therefore, to build and optimize large models and support more generative artificial intelligence applications, it is not only necessary to increase the computing power of the cloud, but also need the cooperation of more powerful computing power at the edge and the end side to form a "end-edge- More balanced computing power distribution under cloud "hybrid computing architecture." To support the operation of personal large models, the intelligent computing capabilities of the terminal must be improved. PC is the most important productivity tool for individuals, and AI PC has become inevitable and necessary to comply with the development trend of large models.

AI PC will have five core characteristics

Yang Yuanqing mentioned in his speech that compared with today’s personal computers, AI PCs in the future will have five core characteristics: first, AI PCs can run personal large models that have been compressed and optimized for performance; second, they have stronger The computing power can support heterogeneous computing including CPU, GPU, and NPU; thirdly, it has larger storage, which can accommodate more personal life-cycle data and form a personal knowledge base to facilitate the learning of personal large models. , training, reasoning, and optimization provide fuel; fourth, it has smoother natural language interaction, and can even use voice and gestures to complete the interaction; fifth, it has more reliable security and privacy protection.

"In other words, your future artificial intelligence personal computer can not only serve as the entrance to a public large model, but also independently run a personalized private large model. It has the most comprehensive personal data and information, and can strictly guard your Its secret. Only you can wake it up and use it, and only it understands you best, far better than the public model." He explained.

For example, if a user wants to make a travel plan, he or she can communicate with his or her computer even in a non-networked airplane mode. The computer can recommend the user's favorites without even telling them their needs and preferences. Flights, hotels, routes and restaurants

" He said: "In the future, your computer, mobile phone and even car will become your personal artificial intelligence twin, like your digital extension and digital mirror, greatly improving your quality of life and work efficiency. ”

The above is the detailed content of The future shape of AI PC: Yang Yuanqing's outlook. For more information, please follow other related articles on the PHP Chinese website!