How does artificial intelligence “need both” and “need”? Yang Yuanqing gives the answer

On October 24, Lenovo Group announced its comprehensive artificial intelligence product technology and the vision of "AI for All" at Tech World.

Yang Yuanqing, chairman and CEO of Lenovo Group, said that for the first time, artificial intelligence technology has become so real and closely related to everyone and every company. While rejoicing in technological progress, we have also begun a deeper exploration.

What “key but intangible” factors support the effectiveness of generative artificial intelligence and large language models? There is no doubt that data is an essential element in the first phase of digital transformation. The training and inference of algorithms and models are key, but they are also inseparable from the support of computing power. Powerful computing capabilities, whether in the cloud, at the edge, or on the device side, are indispensable

To create panoramic artificial intelligence, we need computing power "from pocket to cloud" and various forms of applications, and solutions for different industries are also critical.

There is a dilemma here. If the user wants the public model to have the correct and appropriate content when answering your questions or talking to you, then you must truly tell it what you really think. , real records and data, but in that case, the user's personal data and even privacy as well as the company's business secrets will become part of the public information.

People not only want to have the function of answering questions with large models, but also hope that their data will only be retained on their own devices or within their own enterprises. How can we achieve "both - and also"? The answer is obviously not limited to public large models, we can achieve it through personal large models and enterprise-level large models.

Lenovo’s large model compression technology allows our own smart terminals and devices to have the ability to run personal-level large models. On these terminals and edge devices that support artificial intelligence functions, local knowledge bases will be built to better understand users. Personal large models will use personal data stored on devices or home servers for inference. Users' personal data will never be shared or sent to the public cloud unless authorized by the user, thus ensuring personal privacy and data security.

Enterprise-level large models will coexist with public large models and public clouds, forming a hybrid form and hybrid deployment of artificial intelligence. This hybrid trend has already emerged in many scenarios, such as hybrid offices and digital office spaces.

The above is the detailed content of How does artificial intelligence “need both” and “need”? Yang Yuanqing gives the answer. For more information, please follow other related articles on the PHP Chinese website!