Alibaba, ByteDance, and Baidu Compete in AI Hardware, Selling "Shovels" for Large Models

06/21 2024 347

AI is changing consumer electronics products.

If 2023 was the year of concentrated AI application explosions, then 2024 is the first year of AI hardware. Whether it's domestic tech giants like Honor, Xiaomi, OPPO, vivo, Huawei, Lenovo, or overseas giants like Apple, Microsoft, Qualcomm, Samsung, all are emphasizing the integration of AI large model technology and applications, especially hardware.

From mobile phones to PCs, AR glasses, TVs, refrigerators, cars, various household robots, and various other products, a large number of products are beginning to deeply integrate the latest generative AI technology, providing consumers with a more comfortable user experience. To a certain extent, AI is becoming ubiquitous and omnipotent through these consumer electronics products.

So what AI hardware do you use the most? Do you know what large models it uses? Don't worry, today I'll take you through the large models behind AI hardware.

Large Model Sharing Accelerates Hardware AI Integration

Before starting the inventory, let's first understand this AI technology competition. Currently, many leading internet companies and AI companies both domestically and internationally have released their own large models, and while empowering their own software applications and hardware products, some participants are sharing their large models with more manufacturers to accelerate the AI integration process of their products.

Recently, the moves of ByteDance's Volcano Engine caught my attention, as it jointly launched a Smart Terminal Large Model Alliance with OPPO, vivo, Honor, Xiaomi, Samsung, and ASUS. According to introductions, applications such as OPPO's Xiao Bu Assistant, Honor MagicBook's YOYO Assistant, Xiaomi's Xiao Ai Assistant, and ASUS's DouDing AI Assistant have all accessed the DouBao large model service.

Image source: DouBao

Last month, at the 2024 Spring Volcano Engine Force Conference, ByteDance unveiled the mysterious veil of the DouBao large model family and announced the official launch of external services. Unexpectedly, in just one month, large models such as DouBao General Model Pro, DouBao General Model Lite, and DouBao Role-Playing Model have begun to address industry-wide challenges in language style, content real-time, associative capabilities, and intent recognition.

Some may think that introducing other large models means that manufacturers lack self-research capabilities, which is not something worth bragging about. In fact, manufacturers do this to quickly keep up with industry competitors, provide users with a better user experience, and introducing does not mean abandoning self-research. When the two refer to each other, it can indeed accelerate the AI integration process of hardware products.

AI + Mobile Phones: Large Models Are Connected As Much As Possible

Over the past year, manufacturers such as Huawei, Xiaomi, Apple, Qualcomm, and Lenovo have all announced their "AI + Hardware" plans. Many manufacturers have chosen to "walk on two legs," introducing one or more large models from outside while also conducting independent research and development behind the scenes.

In terms of domestic mobile phone manufacturers, Xiaomi, OPPO, vivo, and Honor have accessed multiple third-party large models. For example, Xiaomi, OPPO, and vivo have accessed Alibaba Tongyi, Baidu Wenxin, and ByteDance DouBao; while Honor has accessed Baidu Wenxin and ByteDance DouBao.

Image source: ByteDance

For foreign mobile phone manufacturers, because the domestic market requires registration before using generative AI technology to provide consumers with generated content, overseas sales products and domestic sales products usually adopt different large model solutions.

For example, the AI functions of the Samsung S24 series are powered by Google's Gemini at the bottom level, while the S24 series sold domestically is supported by Baidu Wenxin's AI capabilities; Apple, on the other hand

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.