Released the first AI PC, teamed up with NVIDIA, what’s different about Lenovo’s new AI wave?

In 2023, when the competition between AI and chips is at its peak, it is not easy to get the full support of NVIDIA and AMD at the same time, but Lenovo is an exception; it is extremely difficult to get Intel and Qualcomm to stand for the same brand, but Lenovo is also an exception.

On October 24, CEOs of NVIDIA, AMD, Intel, Qualcomm, Microsoft and other technology companies rarely attended the same event. It was the 9th Lenovo TechWord – Lenovo Group CEO Yang Yuanqing presented a series of AI-themed products at the event. Products and technology, in Yang Yuanqing’s words:

AI For All.

Lenovo is one of the most popular PC brands in the world. The ThinkPad series alone has sold more than 200 million units. Whether it is an AI chip or AI software, it can only maximize its value when it enters the use scenario. I think this is the Lenovo TechWorld Conference The reason for so much attention.

Ai Faner also came to the scene for the first time . Through communicating with the CEOs of major technology giants, she gradually gained a different understanding of the "new era of AI empowerment" described by Lenovo; she also learned from each company's different views on AI. Through the interpretation of perspective, we have a deeper understanding of "AI for All".

Lenovo’s next-generation computer—AI PC

In fact, "AI For All" is not a new concept. Lenovo had already confirmed this strategy as early as 2017 and continued to promote the group's transformation. Today, Lenovo mentioned "AI For All" again on TechWorld, more in the hope of showing its results to the industry. The most important product is: AI PC.

To be precise, AI PC is not a specific product, but Lenovo's definition of the next generation PC. The actual products on the market will not be delivered until next year.

Luca Rossi, President of Lenovo Intelligent Device Business Group (IDG), also revealed to Aifaner in an interview after the meeting some surprises that AI and PC will bring in the next one or two years:

What I want to share with you is that in the future, our CPU processor will be loaded with an NPU. NPU will help us better manage AI-related workloads, improve performance, reduce battery consumption, and improve overall speed and performance.

At the same time, matching software will be brought to the market, not only from Microsoft, but also from many other independent software vendors (ISVs). These software will make full use of the power of NPU to run AI algorithms and AI reasoning at high speed.

AMD has launched its Rise in AI processor. In the next two to three months, there will also be Intel's Meteor Lake. Both CPUs are loaded with NPU, which can greatly accelerate AI reasoning capabilities.

At the same time, Lenovo Group Chairman and CEO Yang Yuanqing also gave a very honest answer about the approximate time when AI PC will meet users, as well as Lenovo’s predictions for the PC market:

AI PC will definitely not be seen this year, but it will definitely be seen next year. But it is not the most ideal and perfect, and may need to be optimized from generation to generation.

As for the PC market, it will not be stimulated by AI PCs for most of next year, and the market should resume growth next year. If possible, return to year-over-year growth from the beginning of the quarter. The past few quarters have been improving quarter by quarter. We believe that this quarter will not only be better than the last quarter, but also very likely to see positive growth year-on-year.

Next year, the global personal computer market will probably grow in single digits around 5%. Of course, AI PC also has a maturation process. It may be a high-end product at the beginning. After its release, it will be good to account for about 10% of the entire market, and then gradually push it to the mainstream.

But I believe that in the future every computer may be an artificial intelligence computer.

The ability to run large personal models on personal smart terminals will be the most important capability of AI PCs.

Lenovo believes that AI PC devices will be digital extensions of users and "personal artificial intelligence twins." Lenovo uses large model compression technology to enable large models that originally require powerful computing power to be able to handle most tasks locally, reducing reliance on cloud processing capabilities.

But currently, managing edge computing is like managing 10,000 different stores, which is a very difficult thing at least in terms of scale.

How to manage and how to balance providing better computing power with more economical and sustainable development? Kirk Skaugen, President of Lenovo Infrastructure Solutions Business Group (ISG), introduced Lenovo’s technical methods with practical examples:

We developed a Lenovo Open Cloud Automated Deployment Platform (LOC-A) that can respond to customer needs more quickly, which has won us a large number of edge use cases.

For example, if you're a fast food chain, you just verify the store number on the edge side and it will tell you if you have a kiosk, a drive-thru service, etc. Then we can download everything from the cloud to bare metal and run it on Installation is complete within minutes. If it had taken weeks to complete using traditional methods, people would have shipped it to a warehouse and left it there for months, and by the time it was sent back for use, the software system might have expired.

Now with Lenovo's open cloud automated deployment platform, we can join hands with our partners. First, we can provide CPU patches, security patches, and application updates so they are available immediately after the server is deployed. Second, we're a big believer in a hybrid AI world where it would be nice to have part of the stack in the cloud, but if you don't have an internet connection, you still need to run your store, you still want to have analytics, you still want to do all the things like know your employees How many hours were spent working in the store, making sure the kiosks and everything were working.

So you can’t move everything to the cloud, but you should have a balance of local and cloud deployments.

At the same time, local knowledge bases are built on these devices, and personal large models will use personal data stored on the device or home server to perform inferences. On this basis, Lenovo has extended the concept of "enterprise-level artificial intelligence twins" to provide enterprise-level intelligent solutions.

Lenovo's hybrid large model – "lock" data security

Lenovo believes that the four characteristics of continuous learning, providing personalized services and protecting user privacy, equipped with knowledge base to adapt to user needs, and natural interaction will be the key to AI PC.

To create a panoramic artificial intelligence, it is necessary to establish a "personal-public" communication channel, which not only brings convenience to communication, but also creates a dilemma.

People not only hope to have the function of answering questions with a public large model, but also hope that their data will only be retained on their own devices or within their own companies, so that personal privacy and business secrets will not become part of public information.

If you want to achieve "both – and need", Lenovo's answer is: to achieve it through "personal large models" and "enterprise-level large models".

Lenovo's large model compression technology uses personal large models to perform inference on personal data stored on devices or home servers. 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 of artificial intelligence with hybrid deployment.

Simply put, similar to the "offline mode" or "stand-alone mode" of online games, your usage data will only be saved on your computer. Of course, as long as you choose "Don't upload to the cloud", you don't have to worry about personal privacy leaks. question.

As it happens, Microsoft and Lenovo have a lot of cooperation in "hybrid large model" research.

Lenovo is also discussing with Microsoft how to integrate the related vision of personal artificial intelligence twin (AI Twin) and enterprise-level artificial intelligence twin into future solutions.

Microsoft combines Lenovo's devices and services with technology to provide help and support to customers in all walks of life. Both parties will also continue to work hard on the basis of existing cooperation.

When Ai Faner interviewed Kirk Skaugen, President of Lenovo ISG, about Lenovo’s solution bias, he said this:

Lenovo's goal is to provide customers with the integrated services they want. Rather than just advertising how good our hardware is, Lenovo provides complete solutions including hardware, software and services.

Yang Yuanqing and Huang Renxun’s big move: hybrid artificial intelligence solutions are coming

So far, ordinary users are not very aware of large models, but in fact many Fortune 500 companies have deeply integrated large models with their own businesses. Therefore, the outbreak of AI PC is likely to start in the enterprise market.

At this year's TechWord, Lenovo also announced that it will cooperate with NVIDIA to launch a new hybrid artificial intelligence plan to bring next-generation cloud artificial intelligence technology to enterprise users around the world.

Lenovo introduces artificial intelligence-driven computing power to every place where data is generated, from the edge to the cloud, helping enterprises to easily deploy customized generative artificial intelligence applications to promote innovation and transformation, covering thousands of industries.

Each enterprise will have different needs for the construction of large private domain models. For example, customers in manufacturing, transportation, medical, finance, energy, etc. all have different requirements for large models.

Huang Jianheng, President of Lenovo Services Group (SSG), summarized all demands into three aspects – customer service, marketing, and operation and maintenance efficiency:

Now I see that there are about three relatively mature AI scenarios. The first is customer service, especially in places where there are many customers, such as banks, operators, and global operators, because they have large call centers. , originally facing these needs of huge customers. Second, market promotion, in fact, the establishment of AI applications is also very mature. Third, the improvement of operation and maintenance service efficiency, such as using AI App to make our delivery faster and more efficient.

Our own internal research has also shown that in addition to these three scenarios, Gen AI can actually achieve better results in all professional knowledge requirements.

For example, let's say I'm talking to a university. This university may have 50,000 applications for admissions every year. How can you read 50,000 applications one by one? And after reading it, you can select good students for the next step of interviews? This is a very good problem that can be solved by AI.

Another example is lawyers, because many documents actually have rules at the back. How to read the contract and how to draft a contract are also scenarios that are very suitable for using AI.

Yang Yuanqing also gave the latest outlook on the partnership with NVIDIA in the field of artificial intelligence:

NVIDIA has been relatively important in our strategic partnerships in the past and may become even more important in the future.

In the past, our gaming computers used Nvidia graphics cards, and our workstations also used Nvidia's GPUs, as did high-performance computing. We had extensive cooperation in many aspects that required GPUs. We are also an important and large customer of Nvidia, and our cooperation will definitely be more extensive in the future, especially extending to new areas such as AI computing and artificial intelligence infrastructure.

NVIDIA also hopes to not only occupy the market of public large models, but also make more achievements in the fields of enterprise applications, private artificial intelligence, private artificial intelligence based on private clouds, etc., including bringing artificial intelligence computing to the edge. Take it to where the data is generated. NVIDIA is ambitious in these areas.

Of course, NVIDIA provides the chips, but in the end we have to make the products and bring them to the market. Therefore, both of us will be a more comprehensive partner, from terminal equipment to edge computing to data centers, to private clouds and public clouds.

Lenovo’s new wave of AI

In Lenovo's vision, AI PCs in the future are not just PC terminals equipped with AI capabilities, but both can provide users with a smarter future life through a combination of "traditional" and "emerging".

Rui Yong, chief technology officer and senior vice president of Lenovo Group, used three scenarios to connect the relationship between the two:

How can AI help PCs? When you use this mode in different scenarios, it has actually helped you optimize a lot of things. When should you give you the most powerful power, all the calculations, and all the memory.

For example, if you are playing a game, he will know that you are playing and directly adjust the performance to the highest level for you. But if the performance is adjusted to the highest level, it will obviously use a lot of battery. But sometimes I just go online and reply to emails. I don't need to use my CPU and battery at such a high performance, so it will help you allocate processing. For example, when I put it on my lap, I might be surfing the Internet. When I put it on the table, I might be working more seriously. These two will do this kind of differentiated optimization.

How can PC help AI? Just now Yuan Qing mentioned that after the emergence of new large models, the PC must be a carrier of AI. I think in addition to being a carrier, we also think it is also an entrance. In the future, large models must use the device side as an entrance. Lenovo releases more than 100 million new devices every year, including mobile phones, tablets, PCs, etc., which will become the entrance to new personal models.

In the next three years, Lenovo will invest US$1 billion in AI research and development. In Yang Yuanqing's view, "US$1 billion is not too much, but simply not enough."

Just the second week after the Lenovo TechWorld conference, Xiaomi officially released ThePaper OS, which creates a "full ecosystem for people, cars and homes", in which HyperMind, the intelligent thinking center that can cross devices, is impressive. In early November, vivo's self-developed AI large model and operating system will also be released at the vivo Developer Conference. Looking overseas, OpenAI's latest valuation has exceeded US$80 billion, and in less than a year, it has become one of the top three unicorn companies in the world.

The rapid development of the AI ​​field is reshaping the competitive landscape of the global technology industry. As Yang Yuanqing said:

Competitors are different every day. When we first started making PCs, our competitors at that time were domestic brands. Today is completely different, so it is always changing… We must bravely stay on top of the trend, that is, we must become a leading company instead of a catching-up company while grasping the general trend and direction.

# Welcome to follow the official WeChat public account of aifaner: aifaner (WeChat ID: ifanr). More exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo