After OpenAI launched the GPT-4o and became the focus of attention, Google took the initiative and AI's heavyweight products kept flashing, sparking a new round of attention and discussion in the industry. The Financial Investment News reporter noticed that at the press conference on May 15th, AI became the undisputed biggest protagonist. According to Google's statistics, the term "AI" was mentioned 121 times during this approximately 2-hour press conference.
Xing Huanlai, Associate Professor and Doctoral Supervisor of the School of Computer and Artificial Intelligence at Southwest Jiaotong University, stated in an interview with reporters that the competition in the field of AI is becoming increasingly fierce, and the technological characteristics of AI can easily lead to a "winner takes all" market phenomenon. Google's rapid technological upgrades and product updates in the AI field also send a signal to other technology companies: continuous research and development and technological updates are the key to maintaining market position.
Drawing Qing Zixiu
Google AI welcomes the Gemini era
At the press conference, Google announced that it has fully entered the "Gemini era" and released a major upgrade to its large model Gemini, as well as a series of supporting "AI family buckets", covering multiple aspects such as AI models, AI search, AI voice assistants, AI chips, etc.
Specifically, Google has released the latest AI model for the Gemini series, the Gemini 1.5 Flash. Previous models included the Gemini Nano, which can run offline on Android devices, the Gemini Pro, which provides advanced reasoning, planning, and comprehension capabilities, and the Gemini Ultra, which is designed specifically for data centers and enterprise applications.
In terms of performance, taking the upgraded Gemini 1.5 Pro model as an example, its context window will be upgraded from 1 million tokens to 2 million tokens, which is referred to by Google as the longest context window among all basic models to date.
The multi-modal real-time AI assistant Project Astra launched by Google has powerful interactive and multi-modal task processing capabilities. It can receive information through video images, understand complex and ever-changing environments, and respond accordingly. Google's new search engine with generative AI capabilities has also become a hot topic. According to Google, its AI Overview feature allows users to conduct AI driven searches through questioning and chatting, making it one of the biggest updates to its search engine in 25 years. In addition, it has been revealed that Google's upcoming Android 15 system will incorporate a series of AI related features.
In terms of hardware, Google has unveiled its sixth generation TPU chip Trillium. Compared to previous products, Trillium has improved performance by nearly 5 times, and Google may use Nvidia's latest Blackwell architecture GPU early next year.
Despite the superior performance of Google's new models and the abundance of related AI products, some argue that it still lacks innovation and core competitiveness in the fiercely competitive large model market environment.
"Although the Gemini model leads the GPT-4 in 30 out of 32 benchmark tests, its performance and stability in some application scenarios still need to be observed," Xing Huanlai analyzed. "Gemini and Project Astra lack core innovation. Although their performance is relatively excellent, they lack core appeal for most ordinary users. Moreover, in such a fiercely competitive large model market, the improvement in model performance is also somewhat aesthetically tiring."
The strength of the competitors that Google AI products need to face is very strong. Xing Huanlai pointed out that although Project Astra can continuously encode video frames and combine video and voice inputs onto the event timeline, its competitiveness is still insufficient compared to GPT-4o, which has the ability to understand speech emotions.
Further upgrading of industry competition
After a day, two leading AI companies have successively released heavyweight products, and the signs of overlapping functions have naturally become the focus of industry attention.
For example, Google's real-time AI assistant Project Astra is considered by many to be close to OpenAI's GPT-4o positioning, and the former's Wensheng video model Veo, Wensheng image model Imagen 3, and other products have also made many people feel that they have a sense of familiarity with products such as Sora and Dall-E. Even with Google's AI search engine assistant AI Overviews, many people associate it with OpenAI's previously highly anticipated but difficult to produce search engine plan.
In this context, the market's judgment about the increasingly fierce competition in the AI field has also been further validated.
From the situation of OpenAI and Google, Xing Huanlai said that the competition between these two will drive progress in the entire AI field, stimulate the R&D enthusiasm of other enterprises, and the competition will become increasingly fierce.
From the recent release of large-scale model products, it can be seen that the main competitive areas in the industry are constantly expanding, and the application scope is also expanding, resulting in a greater social impact. As a result, it has attracted more and more investors to pay attention to the AI field and continuously increase support, leading to more intense competition among the top players in the AI field.
AI application investment opportunities attract attention
However, intensified competition has also played a role in accelerating the development speed of the industry. In the current field of AI, the constantly improving product performance and increasingly diversified AI applications are all reminding relevant enterprises to continuously improve their product competitiveness and seek new innovative points and service models.
In Xing Huanlai's view, the competition in the field of AI is becoming increasingly fierce, and the technological characteristics of AI can easily lead to a "winner takes all" market phenomenon. Google's rapid technological upgrades and product updates in the AI field also send a signal to other technology companies that continuous research and technological updates are the key to maintaining market position. Nowadays, diversified AI application scenarios are becoming a reality, which also reminds enterprises to find AI application points in various fields, create new business value and service models, and prepare for AI competition.
In fact, some revenue related to AI has become a new growth point for some domestic giants.
Alibaba's performance in the fourth quarter of fiscal year 2024 shows that during the reporting period, Alibaba's AI related revenue accelerated to triple digits, driven by basic model companies, Internet companies, financial services, automotive and other industry customers; Robin Lee, the founder of Baidu, mentioned in an internal speech that at present, 10% of Baidu's large search traffic is generated through ERNIE Bot, and 2.5 million users use the AI function of the library every day.
Guojin Securities believes that the current demand for AI is growing rapidly, and the technology is also in a stage of rapid iteration but not yet mature. The products of leading manufacturers have higher maturity and competitiveness, and the growth rate and certainty of performance are higher. In the long run, AI models are gradually maturing, and the level of intelligence is further improving, which is expected to drive a large number of applications and further drive up the valuation of related companies. In the overseas supply chain, there have been significant upgrades in the quantity and technology of wafer manufacturing, heat dissipation, and storage, which are expected to fully benefit from AI technology iteration and strong demand.
Huatai Securities stated that from an investment perspective, the biggest investment opportunity for the AI big model lies in its application, and it is expected that the big model will be the first to be implemented in industries such as search, e-commerce, retail, office, finance, and medicine.
