Image source: Canva

Since the outbreak of GPT-4 last year, the hype surrounding AI big models in the domestic industry has gradually reached a climax. In the capital market, all companies related to AI have taken turns to increase in value.

However, a careful review of the financial reports of AI's Four Dragons reveals that these once "old timers" in AI still haven't seen much improvement in their fundamentals. Facing the rapidly approaching big models, they generally appear slightly sluggish, even lagging behind some traditional giants and AI startups. But fortunately, AI's Four Dragons have ultimately boarded the "free ride" of AI big models.

AI's Four Dragons "Board" the Big Model "Free Ride"

After the outbreak of foreign AI big models in 2023, China also quickly emerged with a wave of catching up. Baidu's Wenxin Yiyan, iFlytek's Spark, Huawei's Pangu, 360's Zhinao, and Alibaba's Tongyi... All of a sudden, domestic giants flocked in, creating a lively scene.

After the AI big model wave was sparked in China, SenseTime and CloudWalk Technology have both launched their "SenseNova" big model system and "Congrong" big model. Yitu Technology and Megvii, which had been silent for a long time, were questioned for their "difficulties in childbirth." Until November last year and January this year, the latter two's big models finally arrived. Although late, they have still boarded the "free ride" of big models. Looking at the current situation of AI's Four Dragons, actively boarding this "free ride" of AI big models is not only a trend but also an important path for them to seek a breakthrough.

First, AI's Four Dragons have a considerable data foundation in their respective application areas, and the conditions for leveraging industry big models are uniquely favorable. For example, SenseTime has advanced technology and leading advantages in areas such as computer vision, speech recognition, and natural language processing. Therefore, SenseTime's AI big models exhibit the characteristics of being "diverse and interconnected." Based on its own technology, it has formed a series of big models under the "SenseNova" system, such as the language big model "Shangliang," AI-generated images with "Miaohua," 2D/3D digital human generation with "Ruying," and generative AI models and applications for large scenes and small objects such as "Qiongyu" and "Gewu."

Compared to SenseTime, Megvii has rich experience and technical accumulation in the field of face recognition and is gradually shifting from the past AI vision field to the multimodal big model field. Yitu Technology has many innovative and leading technologies in speech recognition and natural language processing, and its big models still focus on traditional business areas such as healthcare and security. CloudWalk Technology excels in providing solutions for smart finance, smart travel, and smart cities, with its AI big models centered around smart cities. In summary, each company has its own strengths, and the directions of their respective big models also vary.

Secondly, big models can accelerate AI's Four Dragons in reshaping their businesses, which is conducive to reducing costs and increasing efficiency. Firstly, with pre-trained big models, the time and cost from user demand to delivery will be greatly compressed. On the one hand, replacing small models with big model technology can greatly compress production cycle time and reduce algorithm development costs, allowing customers to enjoy the benefits of technology faster. On the other hand, end-to-end integration of algorithms and computing power can also lead to cost minimization. For example, in the past, it took 16 cabinets to achieve 10,000 video streams of intelligence, but now with end-to-end integration of algorithms and computing power, only one cabinet is needed to solve all problems, but the overall cost has dropped by 80%.

For AI's Four Dragons, which have been losing money for years, reducing R&D costs and enhancing customer experience have become a top priority. Financial reports show that SenseTime's first-half fiscal year 2023 performance report shows that total revenue was 1.433 billion yuan, a 1.27% increase from 1.415 billion yuan