Technology giants accelerate the layout of AI big models, and competition is rising again

Looking ahead to the future of AI, experts say that most companies can use AI tools to discover optimal operating models and make reliable predictions of production management and market demand and evolution, and AI and its tools will be key to driving the next generation of consumers and enterprises in various industries information technology applications and tools.

Reporter Liu Hui

Artificial intelligence is changing the world, and the speed of change is far greater than people think.

Not to mention that ChatGPT has passed the American physician and certified public accountant exams, it has also entered the homes of ordinary people. One internet mogul said that his elementary school daughter created her first novel using ChatGPT and one-click artificial intelligence (AI) images, and he was very happy after reading it. Another tech company founder said he was practicing spoken English with AI, which can simulate various life scenarios.

Some economists say ChatGPT is becoming the 25th general-purpose technology, after humanity's greatest invention was the wheel. The generative AI large model platform represented by ChatGPT is a milestone tool in the development of AI and a product of the development of information technology revolution to a specific historical stage. Many people will be curious to ask: what technological breakthroughs are global technology companies in the field of artificial intelligence? What is the current situation of artificial intelligence layout of listed companies in China? How will artificial intelligence change the world in the future?

What the world's top AI companies do

The fastest progress in global AI technology continues to come from the United States and Europe. Musk, the founder of OpenAI, which developed ChatGPT, has quit the company, and he said on May 23 of the gap between artificial intelligence between China and the United States, "The gap is about 12 months." ”

Yi Weidong, one of the leaders of China's Internet of Things industry and a professor at the University of Chinese Academy of Sciences, said in an interview with the China Economic Times that in addition to OpenAI, the world's top artificial intelligence companies include NVIDIA, Microsoft, Google, Amazon, and IBM.

Yi Weidong introduced to reporters what global technology giants have been doing recently. NVIDIA is a world-class giant of AI computing power and computing hardware platforms, its GPU, DPU, CPU, systems and their software are in a leading and leading position in the market, creating supercomputers suitable for AI use and solving AI problems that ordinary computers cannot solve, Nvidia AI frameworks and development tools can quickly build, deploy and manage custom applications.

Microsoft is the initiator of this wave of AI, not only investing heavily in AI, but also cooperating with OpenAI and NVIDIA. Microsoft is developing a chip called Athena, with the goal of providing a computing power engine for OpenAI and eventually stealing the market for Nvidia through Microsoft's Azure cloud service.

Google is the leader of the underlying AI model and its application, a global search engine giant, and the leader of this round of AI technology wave. Google has invested heavily in machine learning and artificial intelligence, developing a natural language model BERT with superior human language understanding and has been used in search engines, and recently developed a generative artificial intelligence tool called Dialog Applications Language Model (LaMDA), which uses technology similar to OpenAI.

Amazon is a leader in the application of AI industry, developing artificial intelligence algorithms to support its recommendation engine for e-commerce, video and music streaming, and also using artificial intelligence to determine product rankings and optimize logistics operations. Drone delivery, which has long sought to implement, is another AI application for the tech giant. AI provides a better foundation for companies to enter new businesses quickly and efficiently, and explore new markets.

IBM is a leader in the delivery of AI enterprise solutions and services. IBM focuses on developing and applying artificial intelligence to enhance human intelligence, improve management efficiency or reduce production costs.

Du Xinze, deputy general manager of the artificial intelligence industry research center of CCID Consulting, said in an interview with the China Economic Times that from a global point of view, in addition to OpenAI, Google, Deepmind, and Microsoft, there are three representative algorithm research and development companies: Anthropic, AI21 Labs, and Stability AI.

Anthropic is an AI technology company founded by the former OpenAI team, and Claude launched similar features to ChatGPT.

AI21 Labs is a company founded in Israel in 2017 that provides natural language processing services.

Founded in 2020, Stability AI is mainly engaged in the research and development of generative artificial intelligence, and the Stable Diffusion model launched can create high-resolution image creation, which directly promotes the application and development of AIGC (artificial intelligence content generation) in virtual reality, games, film production and other fields.

In Du Xinze's view, the artificial intelligence technology revolution triggered by ChatGPT is extremely subversive for the industry, on the one hand, the AIGC application based on generative pre-training models breaks the inherent cognition of machine capabilities; On the other hand, advances in machine understanding brought about by big models will disrupt the way humans interact with machines, changing future business models.

Kai-Fu Lee, chairman of Innovation Factory, said, "Although the current AI model is not perfect, it contains tens of trillions of dollars of business opportunities. "But big model training is also only the strength of technology giants." Dai Qionghai, chairman of the Chinese Institute of Industrial Intelligence, said that the cost of algorithm large models is becoming increasingly high, with GPT-3 training costs of about $5 million per time, and GPT series research investment exceeding $3 billion. And the model is getting bigger and bigger, and the GPT-3 training data is about 450,000 years of text volume in People's Daily.

Without sufficient investment, it will not be possible to directly participate in the next generation of AI industry

There are many concept stocks related to artificial intelligence in China, such as algorithm models such as Baidu and 360, supercomputing power such as Zhongke Sugon, Huawei, Inspur Information, and basic computing power such as Aofei Data; "AI+Education" such as iFLYTEK, "AI+Industry" such as Central Control Technology, "AI+Healthcare" such as Entrepreneurial Wellcom, "AI+Finance" such as Huishun, Oriental Wealth, "AI+Cybersecurity" such as Anheng Information, and so on.

Chen Xudong, chairman of IBM Greater China, recently said: "Nearly 60% of Chinese companies are actively deploying AI, and 70% are considering using AI to accelerate environmental, social and governance processes. ”

Yang Chang, chief analyst of Zhongtai Securities Research Institute, told this reporter that the concept of artificial intelligence was proposed as early as the 50s of the 20th century, but due to theoretical limitations and several troughs, since the deep learning algorithm was proposed in 2006, artificial intelligence ushered in the third wave of development. At present, foreign countries represented by ChatGPT-4 have formed a leading advantage in artificial intelligence large models. In China, there are many types of large models such as Baidu Wenxin, Ali Dharma Academy multimodal M6, Pangu large model jointly developed by Huawei and Pengcheng Lab, and Wudao 2.0 of Beijing KLCII Artificial Intelligence Research Institute.

Reporters contacted Baidu, Huawei, Tencent and other companies for interviews about artificial intelligence, but the company failed to give more responses. But Baidu Chairman Robin Li said at the Zhongguancun Forum that "all future applications will be developed based on the big model and will be integrated into the real economy." Big models will change the world. ”

Yi Weidong told this reporter that domestic listed companies involved in artificial intelligence large models and generative AI-related technology sectors basically turned in after seeing the opportunities and popularity of large models, and some companies did not have substantial innovation or breakthroughs in technology. From 2010 to 2018, AI unicorn listed companies emerged in the wave of AI entrepreneurship, and their core technologies are mainly based on small models of deep learning, and their industry applications are limited. As for the domestic listed companies that can really provide computing power hardware for AI large models, they have not yet seen it. In general, the underlying core technologies of these AI-related listed companies are mainly obtained through learning, introduction and digestion, and it will take time to truly own their own unique core technologies.

Yi Weidong said that in terms of the core technology of the original underlying large model similar to OpenAI, Li Yanhong once said: "China will basically not produce another OpenAI." "These companies basically do language or image vision applications, the most are companies that do multimodal applications, and few companies do the underlying large model or large model infrastructure optimization. In the specific direction, a small number of companies are in the field of chatBot, and most of them are in the field of productivity tools, including copywriting, script generation, avatars, video ad generation, etc.

Du Xinze told this reporter that the listed companies in the domestic artificial intelligence sector are mainly divided into three categories: the first is large-scale technology enterprises, the second is the rapid development of enterprises relying on the wave of artificial intelligence, and the third is traditional industry enterprises, relying on the market advantages of specific industries to carry out artificial intelligence research and promote their own business upgrading. The emergence of ChatGPT not only allows people to see the machine creation ability beyond human imagination, but also allows practitioners to see the possibility of artificial intelligence technology transformation from perceptual intelligence to cognitive intelligence, which directly promotes the technological revolution of the entire artificial intelligence industry, and many enterprises have begun to enter the technical competition of ultra-large-scale pre-training models.

Therefore, in Du Xinze's view, to judge whether the company really has technical strength in the field of artificial intelligence in the future, we must first look at how many resources the company will invest in the field of large models, large-scale computing power investment and the entry of more than 100 core algorithm engineers will be necessary conditions for the development of core technical capabilities of large models, and artificial intelligence sector enterprises without sufficient investment in research and development and without the establishment of a special AI algorithm engineer team will not be direct participants in the next generation of artificial intelligence industry.

Looking forward to the future AI prospects, Yi Weidong said that most companies can use AI tools to discover the optimal mode of operation, and make reliable predictions of production management and market demand and evolution, AI and its tools will be the key to driving the next generation of consumers and enterprises in various industries information technology applications and tools, which is very similar to the birth and development of mobile communication and Internet ecosystems more than 10 years ago. Generative artificial intelligence, represented by ChatGPT, will help enterprises bring AI technology to various mainstream areas of the market.