Intel launches the latest Core processors to promote the development of AI PCs, but native AI applications are still insufficient, and the construction of an industrial chain ecosystem is still in its infancy.

Yesterday, the first batch of terminal products equipped with Intel Core Ultra processors went on sale in China.

Gao Yu, technical general manager of Intel China, introduced at the press conference that Intel Core Ultra processors are capable of handling large language models with up to 20 billion parameters. Currently, they have cooperated with Baichuan Intelligent, Wenxinyiyan, and Zhipu in China. Adapt and cooperate with large models such as AI, WPS AI, Tongyi Qianwen, OPPO AndesGPT.

IDC predicts that terminal devices equipped with AI functions will exceed 70% in the Chinese market in 2024, and AI terminals will account for 55%. The assembly proportion of AI PCs in new machines in the Chinese PC market will rise rapidly in the next few years, reaching 81% in 2027, becoming the mainstream of the PC market.

Gao Yu said in an interview after the meeting that he is optimistic that NPU-based application scenarios and AI imaging applications will develop rapidly. However, he also admitted that the AI ​​PC ecosystem has just begun to be built and it will still take a long time to improve the industry chain.

▍NPU-based application scenarios will develop the fastest

Currently, both Intel and AMD have launched processor solutions with built-in NPU units, and Qualcomm has also released built-in NPU processors based on ARM architecture

Many industry analyzes have shown that the "CPU NPU GPU" collaborative architecture will become the computing power base of AI PC in the future. Among them, NPU is suitable for continuous AI and sharing AI load, and CPU is suitable for low-latency AI tasks of single inference

Gao Yu said that compared with calling large models in the cloud, the advantage of AI PC is that it does not rely on the network, is highly customized for individual scenarios, and provides AI services for thousands of people.

Huang Yiqun, general manager of Microsoft's Asia-Pacific Equipment Partner Sales Department, introduced the effect of Windows Conference Room software after applying NPU: online meetings can realize automatic framing, including blurring the background of the meeting and automatically positioning the portrait in the middle of the screen wait.

The XPU architecture is expected to promote the rapid popularization of AI digital humans. Gao Yu said that through the collaborative work of NPU, CPU and GPU, the NPU is responsible for human body motion capture, the CPU is responsible for logical calculation, and the GPU is responsible for three-dimensional rendering, which can quickly generate real digital people on PC

The Beijing Academy of Fine Arts demonstrated their ability to create pictures from pictures through AI PC in a live demonstration, generating Qi Baishi-style paintings in real time

In addition, Intel also cooperated with Jianying to localize the deployment of AIGC models. In the live demonstration, the clipping optimized the "intelligent keying" function through the Core Ultra platform, which can shorten the time by up to 2/3

Talking about the application scenarios of AI PC in 2024, Gao Yu believes that NPU-based application scenarios will develop the fastest. "For example, in editing, using NPU to cut out images, the power consumption is directly reduced by a large proportion, and the time is shortened by 1/3. This kind of application scenario combined with NPU is definitely the fastest to be implemented. ”

Secondly,

AI mapping applications similar to Stable Diffusion will also develop rapidly. "We can now quickly generate images in 1 second and high-definition images in 6 seconds on PC. This speed is completely comparable to the cloud. I think text-generated images and image-generated images will develop faster. And the large language model of pure text interaction Still need to explore, maybe a little slower.”

Compared with the booming development of mobile phone applications, the development of PC in the application layer has been relatively slow in the past many years, but Gao Yu believes that AI PC will change this status quo.

"Domestic Windows applications are not as blooming as on mobile phones, and the difference is two orders of magnitude. Now, to a certain extent, AI PC has created demand and lowered the development threshold. So you should see Many small software vendors are combining it with AI on PCs to create some new things. At this time, application stores may be needed to publish applications, which will most likely appear in the near future."

AI PC ecosystem is under active construction

Gao Yu believes that although the war of hundreds of models has begun in China, AI native applications developed based on models are still very limited. He believes that this is because the industrial chain ecology is still in the construction stage

Although early artificial intelligence applications are not as popular as they are today, we must not be limited by thinking. Artificial intelligence is changing very rapidly. We could not have imagined today's situation a year ago, and it was even more unimaginable two years ago. So, can we imagine what it will be like in a year, two years? But the predictable trend is that the entire industry is developing in this direction

At the press conference,

Intel announced Core Ultra processor hardware and channel partners, including Asus, Acer, Dell, H3C, Honor, HP, Lenovo, Xiaomi, JD.com, Pinduoduo, Suning.com , Tmall, etc.; software and model partners include Microsoft, Kirin Software, Tencent Conference, NetEase Youdao, iQiyi, Jianying, etc.

Gao Yu said frankly that

the AI ​​PC ecosystem has just begun to be built and it will still take a long time, but once the ecosystem is established, the situation may be completely unimaginable.

We are still exploring it. This concept is not new and is in the stage of rapid evolution. The first thing we need to do is fit large models, including language models and generative models. This is the most essential capability of AI on a personal computer, allowing it to run smoothly on the computer with minimal loss of accuracy. At the same time, we also need to minimize memory and GPU usage, which requires a lot of optimization work. Once these problems are solved, the industry will iterate rapidly and produce many applications that we cannot imagine today

Rewritten content: Source: Science and Technology Innovation Board Daily

The above is the detailed content of Intel launches the latest Core processors to promote the development of AI PCs, but native AI applications are still insufficient, and the construction of an industrial chain ecosystem is still in its infancy.. For more information, please follow other related articles on the PHP Chinese website!