Recently, technology giants such as Intel, Lenovo, and Qualcomm have made AI PCs (smart computers, including laptops) increasingly popular.
Moreover, even Apple, which has a relatively cautious attitude towards AI, is strengthening the concept of AI PC. At the latest iMac and MacBook Pro product launch on October 31st, Apple announced that its self-developed M3 series of chips can be used for machine learning (ML) models. The M3 Max with up to 128GB of unified memory is sufficient for developers to handle AI large models containing billions of parameters.
In the eyes of the outside world, with the release of the new Mac equipped with the M3 series of chips, Apple has entered the "Vanity Fair" of AI PCs. Is AI PC a reshaping of the PC industry or a gimmick? What new experiences can it bring to consumers? Through this press conference, we can see in what areas of AI does Apple have ambitions?
Regarding questions such as whether Apple is developing an AI big model and future plans for AI, China Business Daily contacted Apple China and has not received a response as of press release.
In fact, running AI on desktop PCs is already a routine operation, and Intel has been promoting the use of generative AI on its Core processors in recent months. Apple's goal may be more ambitious, such as the Max level GPU, which should be prepared to compete with NVIDIA RTX4080 level dominance, "said senior industry analyst Huang Yefeng.
Dada, the founder of Notebook Evaluation Network, believes that AI is currently a disadvantage for PCs, as many AI models are running in the cloud rather than locally, and personal computers are purely "thin terminals". At present, there has been no change. Each AI PC is complete enough to run SD (Stable Diffusion), and the LLM (Large Language Model) will not be launched until 2024. It is still a marketing gimmick to boast about AI with current products. The significance of AI in the current stage for consumers is still to use it in the cloud to improve work efficiency, "said Dada.
Generative AI is expected to become a watershed in the industry
At present, there is a lot of discussion about AI PCs in the industry. For example, NVIDIA CEO Huang Renxun stated that the computer industry is facing a regeneration opportunity, and new AI computers will replace valuable traditional computers in the next decade.
At the Innovation 2023 conference held in September, Intel announced the launch of the AI PC acceleration program. It is reported that this acceleration plan connects Intel's resources, including AI toolchain, hardware, design resources, and technology, for independent hardware vendors (IHVs) and independent software vendors (ISVs).
According to the plan, Intel will bring artificial intelligence features to over 1 million PCs by 2025, and will be driven by the IntelCore Ultra processor released in December this year.
On October 24th, Lenovo teamed up with technology giants such as NVIDIA, Microsoft, AMD, and Qualcomm to discuss "AI for All", focusing on how to unleash the power of artificial intelligence and promote intelligent transformation in various aspects of life and industries; In addition, Lenovo also showcased its first AI PC product, which can create personalized local knowledge bases and use model compression technology to run personal large models, achieving natural AI interaction.
Lenovo believes that AI PC is a brand new intelligent productivity tool tailored for everyone, which will further improve productivity, simplify workflow, and protect personal privacy data security. The future personal computer will be an artificial intelligence computer, the future mobile phone will be an artificial intelligence phone, and the future workstation will be an artificial intelligence workstation. Intelligent devices are the digital extension of users, just like their twins, we call them 'personal artificial intelligence twins'. "said Yang Yuanqing, Chairman and CEO of Lenovo Group.
On October 25th, Qualcomm launched the Snapdragon X Elite, the company's strongest computing processor ever built for PCs, at the Snapdragon Summit. It integrates a newly customized Oryon CPU, Adreno GPU, AI engine, and more, and is a PC SoC (system level chip) with a 4-nanometer ARM architecture. It is reported that OEM manufacturers are expected to launch PCs equipped with Snapdragon X Elite in mid-2024.
Qualcomm introduced that Snapdragon XElite is specifically designed for AI, and X Elite can support running 13 billion parameter large models on terminals, generating 30 tokens per second for 7 billion parameter large language models.
When it comes to AI PC on Apple's side, we have to mention its M3 series (M3, M3 Pro, and M3 Max) processors. In addition, the Apple Mac product line, equipped with the latest M3 series chips, also emphasizes its own accelerated machine learning model. Apple claims that the M3 Max chip with up to 128GB of unified memory is sufficient for developers to handle AI large models containing billions of parameters.
Why have domestic and foreign manufacturers continued to focus on AI PCs in the second half of the year? Zhang Xiaorong, President of the Deep Science and Technology Research Institute, said, "The reason why major manufacturers emphasize AI in PC products is mainly driven by applications. Generative AI software has preliminary application value and needs to be promoted with supporting hardware
According to data from market research firm IDC, global PC shipments increased by 11% month on month in the third quarter of 2023, with shipments of 68.2 million units. Lenovo, HP, and Dell respectively accounted for the top three brand shipments. Generative AI may become a watershed in the PC industry, "said Linn Huang, Vice President of Device and Display Research at IDC." Although the case is not yet fully clear, market interest in this category is already strong. AI PCs can personalize user experiences at a deeper level while protecting data privacy
The global PC industry enters the AI era in the next five years
However, in the view of He Tianqi, former product planning manager at Lenovo, the current significance of AI PC in reshaping the PC industry is not significant. The current concept of locally deployed AI models is a gimmick, but the newly added NPU (Neural Processing Unit) on hardware may have other uses and benefits in the future
Dada also believes that AI is currently a bearish option for PCs, as many AI models run in the cloud rather than locally, and personal computers are purely 'thin terminals'. The next golden cycle for PCs is when companies can massively run Stable Diffusion and large language models on personal computers through significant upgrades to local computing power, "he said. Of course, in addition to the efforts of PC manufacturers OEM and chip manufacturers themselves, it also depends on the performance of the open source model.
Each AI PC is fully capable of running SD and LLM, and it will not be launched until 2024. Using current products to promote AI is a marketing gimmick. At present, the significance of AI for consumers is still through cloud calling, but it integrates user workflows, especially ChatGPT assisted copywriting, code writing, and SD graphics, which improves work efficiency, "said Dada.
Dada also mentioned that next year's Intel MeteorLake processor, also known as the Core Ultra, will provide a complete software and hardware ecosystem, from software support to dedicated NPUs on hardware, which can directly run large language models locally. Although the AMD Ruilong Zen 4 processors currently on the market also claim to support AI, they actually have an NPU hardware, and the software package has not kept up. Except for the video effects that are useful in Microsoft Studio Effects, they cannot be called in other scenarios, "said Dada.
However, Huang Yefeng believes that Intel's software stack in the end-to-end AI ecosystem has begun to take shape. For AI application development on Windows, there are a variety of APIs available, including Microsoft's WinML and lower level DirectML; you can also choose open-source ONNX RT, as well as Intel's OpenVINO. "He took the collaboration between Intel and Microsoft as an example, and Intel launched Windows Studio Effects in the upper application layer - equivalent to a virtual camera, which performs AI processing on videos from the original camera, which can be used for background replacement, blurring, and Face tracking, etc.
For application developers, they only need to call Windows Studio Effects to directly achieve these effects. For example, Microsoft Teams directly calls Windows Studio Effects, and these AI special effects implementations can run on NPUs, "said Huang Yefeng.
We currently have over 100 partners working on various terminal side AI applications, enriching the user experience of PC users. We feel that the development of terminal side AI is still very fast, "Intel said.
Qunzhi Consulting predicts that 2024 will be the first year for large-scale AI PC shipments, and it is expected that the global AI PC shipment volume will reach about 13 million units by 2024. From 2025 to 2026, the AI PC shipment volume will continue to maintain a double-digit annual growth rate and become the mainstream PC product type in 2027. This means that the global PC industry will steadily enter the AI era in the next five years.
However, an observer who declined to be named believes that the application of AI PC is not easy, and even after Apple has been working on NPU for so long, the AI ecosystem on the Mac end is quite inadequate. Although there are now applications such as video conferencing to replace backgrounds, image superresolution, audio noise reduction, and focusing on cameras, the computing power of thin and lightweight laptops is not as high. PC manufacturers are generally talking about the value of generative AI. I think for individual users, generative AI like ChatGPT is definitely more reliable than cloud based services, and local reasoning is not necessary. The value of AI PCs should still be in industry applications, which is beneficial for enterprises Households have value The above observer said.
Apple invests $1 billion annually to catch up with the AI trend
Generally speaking, the improvement of GPU performance is beneficial for improving the computing speed and power of AI, making AI technology more practical and efficient. This also means that Macs equipped with M3 series chips can bring better performance and response speed to AI applications.
Apple emphasized their neural engine at the meeting, with the latest M3 series chip having a neural engine 60% faster than the M1 series chip. This will be very helpful for their future AI based applications, "said Su Lianjie, chief analyst of Omdia AI industry at research firm Omdia.
For AI, Apple was not the first to apply it to its own products, but AI gives Apple products more advantages. For example, Siri voice assistant, facial recognition technology, etc. are typical applications of Apple in the AI field.
Apple's efforts in AI have always been moving towards mobile applications. From early Siri to later computational photography and artificial intelligence inference chips on mobile PCs, Apple is leading the market. This has also inspired many companies to make efforts in the same field, such as Google's Voice AI and Qualcomm's artificial intelligence inference chips, which have a trend of coming from behind. In contrast, Apple appears to be in the iteration of artificial intelligence products Be cautious, especially Siri Su Lianjie said.
Apple tends to lean towards developers in terms of AI ecosystem construction, especially developers of peripheral devices and apps. "Su Lianjie said that Microsoft and Intel, as leaders in open communities, have the style of" brothers within the world, "and are not shy or sidelined. On the contrary, Apple focuses more on its own device ecosystem, focusing on the developer community related to its own devices.
However, AI algorithm expert Huang Song analyzed that this is because Apple places more emphasis on integrating AI functionality into its expertise in the field of consumer electronics software and hardware.
It is understood that Apple parted ways with Intel in 2020 and began using the self-developed chip M-series on MacBook and iMac product lines. Data from IDC shows that since then, Apple's Mac business has experienced a recovery, with market share roughly doubling to nearly 11%. Apple's financial report shows that Mac's revenue in the fiscal year 2022 reached $40.18 billion, a year-on-year increase of 14%, accounting for approximately 11% of its revenue.
According to foreign technology media MacRumors, Apple plans to invest $1 billion annually in the development of generative AI products to bridge the gap with major competitors such as Google, Microsoft, and Amazon. The disclosure also shows that Apple has developed its own large-scale language model Ajax and has launched a chat robot called "Apple GPT" internally.