Lenovo opens a new year for AI PC.In the near future, every individual and business will have their own AI assistant. Since the advent of ChatGPT last year, new generative AI applications have been born every day, and they are "bombarding" people's eyes with cool demos. The rapid development of artificial intelligence has given rise to an explosion in demand for computing power, and computing power has now become a necessity of life in the AI era, just like "water, electricity and coal".
In the face of such a menacing wave of AI, the public seems to be a little confused: these apps sound very high-end, but how do they relate to our daily lives?Ordinary people don't understand AI large models, can they easily get started with these applications?How does computing power bring innovative experience to people?
To truly realize the inclusiveness of AI, it is necessary to overcome multiple challenges such as AI personalized services, efficiency, cost, and security. But it's time to face these difficulties, let AI benefit our future world, and let AI help solve problems that can't be solved today. Wang Chuandong, vice president of Lenovo Group and chief marketing officer of China, said in a recent speech that the future world will be a world full of intelligent body connections, and in the near future, every enterprise and individual will have an intelligent body driven by artificial intelligence. He also believes that having an agent is like having a dedicated assistant dedicated to helping individuals or businesses with various tasks, and the user only needs to tell the agent what you want to do in everyday language, and the agent can make a personalized response.
Wang Chuandong, Vice President of Lenovo Group and Chief Marketing Officer of Lenovo China.
It can be seen that Lenovo, as a PC leader, has been moving continuously since October, and is committed to using AI PCs as terminals to create a full-stack intelligent layout. From Wang Chuandong's speech, as well as the recently held first AI PC Industry Innovation Forum and the newly released AI PC Industry (China)**, we may be able to find a real solution for the implementation of AI.
It's time for AI capabilities to get out of the ivory tower and into public life.
How far is it from AI inclusiveness?
It can be said that the current "emergence" of generative AI is the common result of the maturity of various software and hardware technologies.
In terms of hardware, the rapid development of GPU technology and the continuous updating of the framework have transformed from the role of a "graphics card" that runs 3D images faster than CPUs, and has become a general-purpose acceleration chip with high-performance computing capabilities, which is the main computing power for large model training and inferenceThe development and popularization of cloud computing has fully decoupled software and hardware, and large model developers can call large-scale computing power, storage, and network resources at a relatively low cost, and the development process has been greatly acceleratedAt the same time, algorithms such as Transformer, Alexnet, and Resnet continue to appear, providing more ideas and possible ...... for the tuning and combination of large models
However, when there is no obstacle to the implementation of technology, and applications for diverse scenarios such as Runaway, Pika, and ChatGPT have emerged, why does generative AI still seem to be full of distance from our lives?
On the one hand, there is no common user-friendly interaction model for generative AI applications. Applications such as Midjourney and Pika need to be completed through the Discord forum interface, and due to the large number of parameters of their models, the parameter tuning instructions are more complex and diverse, and most users have programming experience or have enough time to learn to use the application.
On the other hand, current generative AI is trained on public domain data, which is a general data model, which is not "tailor-made" enough. For enterprises, AI models are needed to be able to create real benefits in line with their own production realityFor individuals, it is hoped that the large model can better "understand" their own language habits and usage requirements. Therefore, the "easy-to-use" large model must have the ability to be trained specifically according to the enterprise and personal data to meet the personalized needs of users. At the same time, as the importance of data security becomes a consensus, the fine-tuning ability of large models must be based on ensuring the privacy and data security of enterprises and individuals.
Lenovo AI PC Industry (China)** also pointed out that users not only need public large model services, but also need their own personal large models. Therefore, how to deliver generative AI and what kind of carrier to deliver it in order to take into account cost, user experience, and personalization has become the top issue on the road to AI inclusion.
Terminal manufacturers integrate and innovate, and PCs make AI accessible to everyone
In the first half of the year, the surging wave of computing power was on the supply side, while the AI terminals that were currently intensively landed were on the demand side. Fang Jing, chief analyst of Minsheng's ** electronics industry, said at the forum of "AI New Ecology and New Year of Wisdom".
The current hot AI market has a feeling of "no AI, no products", but which of the various terminals such as mobile phones, PCs, tablets, and smart wearable devices can maximize the release of AI capabilities?
First of all, we must clarify what characteristics a qualified AI terminal needs to have.
From the perspective of computing power, the training task of trillion-level parameter large models still needs to be completed in the cloud, and only strong inference capabilities are required on the terminal side to quickly respond to user needs, and when invoking the cloud model for inference, the terminal must ensure the security of users' local data.
Mobile phones, tablets, and smart wearables are highly portable and enable users to apply AI capabilities anytime, anywhere. However, "small size" is a double-edged sword, which inevitably means that it cannot be built-in with large high-computing power chips, so it is difficult to carry the inference task of large models.
If you consider putting all inference tasks in the cloud with powerful computing power, on the one hand, you must accept the high cost of computing in the cloud, and on the other hand, you need to establish a reasonable mechanism to protect users' private data from leakage. In addition, because the inference results still need to be transmitted back to the terminal output, this mode also puts forward high requirements for network conditions, and it may take tens of seconds to obtain the inference results in poor network conditions, which greatly affects the application experience.
According to the AI PC Industry (China)** released by Lenovo, AI PC will not only be a hardware device, but a hybrid that includes AI models and applications, as well as hardware devices. To put it simply, compared with other terminals, AI PC terminals have full-modal human-computer interaction conditions, are full-scene productivity platforms, the strongest personal computing platforms, and the largest capacity security terminals, standing out among all types of terminals and becoming the preferred terminal for carrying AI computing power.
At the same time, many links in the AI industry chain are constantly evolving: large AI models need to be continuously iterated, computing power needs to be further matched with model requirements, and application ecology needs to be continuously improved. However, ordinary users do not have the ability to control each link and simply want to obtain the best AI application results, so the terminal needs to be able to deliver AI capabilities in a coordinated manner.
Since the development of PC in the early 80s of the last century, it is more complex and more complete and mature than other terminal industry chains. Taking Lenovo Group, a leader in the PC industry, as an example, it has nearly 40 years of experience in the PC industry chain, and has led the progress of the PC industry chain in both miniaturization and network transformation. It can be said that PC manufacturers, as integrators of the previous generation of PC technology, are actually more capable of becoming technology integration innovation deliverers and software and hardware capability integrators in the AI era.
As Amu, Vice President of Lenovo Group and Vice President of Strategy and Business Development in China, said, "With its excellent performance in all aspects, PC has once again become the preferred terminal for AI to benefit everyone, and has undertaken the historical mission of bringing users an epoch-making new AI experience." ”
The time and place are favorable, and hard power is king
Although more than 20 years have passed since the PC has been greatly upgraded to the Internet wave, Lenovo has already taken the top spot in the world's PC shipments because of the first Internet computer.
Considering that the entire industry ecosystem has a mature process, AI PCs will first go through an AI ready stage, and the chip architecture will be the first to upgrade to a hybrid AI architecture and accelerate AI application innovationWith the embedding of large personal models, personal agents, and open ecosystems for AI applications, AI PCs will accelerate into the AI ON stage. Amu, vice president of Lenovo Group and chief strategy officer of China, said.
The PC terminal manufacturers represented by Lenovo are constantly promoting the function upgrade of AI PCs around the goal of providing personalized services in general scenarios, instant and reliable service response, lower cost of using large models, and credible and secure personal data and privacy protection. It is understood that on December 15th, Lenovo released ThinkPad X1 Carbon AI and Lenovo Xiaoxin Pro 16 AI Core Edition two AI ready AI PC products, and they were officially listed at 15:00 on the same day, and the pre-sale was opened at the same time. These two products have three major features: embedded hybrid AI computing power, innovative enhanced AI experience, and device experience upgrade, which means that Lenovo Group's AI PC has officially entered the AI ready stage.
The potential energy of AIPC is enough to be visible, first of all, AI PC has the most modal interaction functions, enriching the interaction mode between users and AI applications. Users can use voice, touch, text, gestures, and other forms to have natural language conversations with computers, not just specific commands or syntax.
Second, AI PCs are equipped with heterogeneous hybrid computing power (CPU + NPU + GPU) as standard, and traditional CPUs for general-purpose computing have AI acceleration capabilities through architectural innovation and integration of NPUs optimized for neural network workloadsIn addition to accelerating game rendering, GPUs also have the ability to perform large-scale parallel data computing. A variety of computing chips can be used together to meet the needs of AI workloads of various shapes and sizes.
At the same time, AI PCs can directly embed personal large models. With computing power, storage, and network resources, AI PCs can deploy a hybrid large-scale model solution that focuses on local and supplemented by edge and cloud. For most localized inference tasks that can be used offline, AI PCs have sufficient and secure storage space to quickly deploy modelsFor a small number of complex inference tasks that must be completed at the edge and in the cloud, AI PCs also have powerful network performance to quickly transmit data and inference results. More importantly, AI PCs can dynamically adjust computing loads on devices and clouds to achieve the lowest cost, best performance, and fastest response speed.
However, because the hybrid large model architecture requires the resources of the cloud, edge, and end to work together, there is a risk of leakage of personal data and enterprise trade secrets. In the face of security performance issues, AI PCs provide device-level personal data and privacy security protection, realizing that software and hardware can be grasped with both hands, and both hands must be hard. At the hardware level, you can configure a hardware-level encryption chip to ensure that the data can be read through the authorization programAt the software level, AI PC can establish a personal data security domain based on local data, and carry out targeted authorization management through the defined data confidentiality level, so that users can enjoy device-level personal data and privacy security protection.
Corresponding to the hard power of PC to carry AI, the market has also given sufficient affirmation to the development potential of AI PC. IDC**, AI PC in China's PC market in the proportion of new machine assembly will rise rapidly in the next few years, will reach 81% by 2027, becoming the mainstream of the PC market.
However, in the long run, AI PCs still need to progress together with AI technology, and the current AI PCs are not perfect, and there is still a lot of room for improvement: it is necessary to explore the way in which AI native applications provide services, whether to form applications independently or "plug-in" calls, and coexist with third-party applicationsThe fine-tuning capability of large AI models still needs to be further developedWhen users need AI to complete complex tasks, they may need to call different applications and models to achieve cooperation ...... between different AI native applications
These changes are not something that can be achieved by a single manufacturer in the industrial chain, but requires an open industry ecology and unified industry standards as support to make the ecosystem prosper together.
As a leader in the PC industry, Lenovo has completed the most important "ecological power" link for the AI PC industry. As the PC brand closest to first-tier consumers, Lenovo was the first to perceive the unstoppable momentum of AI development and the necessity of building an AI PC ecosystem, and shouldered the responsibility of being the standard setter and promoter of an open AI application ecosystem.
At the AI PC Industry Innovation Forum, Wang Chuandong, Vice President of Lenovo Group and Chief Marketing Officer of China, released the "Lenovo AI PC Pioneer Action", inviting more software and hardware partners to join and jointly promote the development of AI PC.
As a leading full-stack intelligent manufacturer and the world's No. 1 PC brand, Lenovo will take the lead in launching the Lenovo AI PC Pioneer Action, and jointly launch the 'Lenovo AI PC Together' series of activities with AI PC ecological pioneers and creative pioneers, so as to live up to AI and go to the mountains and seas together. ”