The explosion of artificial intelligence models, the cloud computing industry ushered in a reshuffle?
Weak human positions will be replaced, and the phenomenon of oligarchy will intensify
Announcement or official debut, including Baidu Wenxin Yiyan, Alibaba Tongyi Thousand Questions, old Internet faces such as HKUST Xunfei, Huawei, 360, Netease, Tencent, Kunlun Wanwei, Daguan Data, Shangtang Technology and even Fudan University and other upstream or academic institutions joined it. For a time, artificial intelligence models sprung up like bamboo shoots after a spring rain, and the scenery was unparalleled. But we know that the foundation of the big model of artificial intelligence lies in cloud computing. Without Microsoft Azure, it would be difficult for OpenAI to develop to the level it is today. It also means that the current cloud computing industry will usher in a wave of inevitable reshuffle and upgrade.
Counting the history of cloud computing in detail, it is not difficult to find that it was born based on the processing of massive data, such as the huge transaction volume of the e-commerce platform’s Double Eleven event, the rush to buy train tickets during the Spring Festival, and other services with explosive traffic characteristics. The server aggregates and classifies the processed data to form targeted group labels, such as what young people like to buy, what old people like to eat, and what girls like to wear, thus forming a big data industry. With the support of big data, the network platform can put different labels on different users, and push the corresponding product advertisements to the eyes of users, which also forms the algorithm recommendation, and the intermediate process of forming this set of links, It is the data analyst who plays a key role. The rapid iteration of large artificial intelligence models means that weak human positions like data analysis will be quickly replaced by highly efficient artificial intelligence. It can not only label users, but also directly tailor them to individuals. Customization, so that the information received by each person is unique, this is the ideal form of big data, and it is also a height beyond the reach of traditional cloud computing.
For enterprises, the fundamental appeal of cloud computing is to reduce costs and increase efficiency. If artificial intelligence can replace some positions, such as customer service, only a small number of senior programmers who maintain artificial intelligence can achieve it. A fundamental impact of the model on the cloud computing industry.
Of course, the current artificial intelligence large language model also has the problem of unbalanced computing power. For example, the performance in the professional field and the strong artificial field is not very good. In our previous tests on Wenxinyiyan, Tongyiqianwen and ChatGPT Here, they will make some very subtle mistakes in the professional field, and even mislead users. Therefore, there will be a relatively strong demand for fine-tuning of large language models in professional subdivision fields. In this scenario, applications such as consulting, customer service, and interactive robots in subdivision fields will explode, and these applications have a high degree of consistency in the technology stack. It is very suitable for the implementation and deployment of SaaS (software as a service), but it also brings certain challenges to cloud artificial intelligence computing power. If everyone uses GPU for training and reasoning, the cost is undoubtedly a big challenge. This is what Musk lamented ” Everyone is buying GPUs”, and precisely because of the high cost, it is basically a large platform that can play large artificial intelligence models, which will undoubtedly promote cloud computing, like artificial intelligence, to further concentrate on top companies.
From to B to to C, cloud computing “expansion” is imminent
Before this, cloud computing was a thing that was almost indifferent to the majority of users although it was ubiquitous. With the continuous evolution of artificial intelligence models, cloud computing will highlight a clearer outline on the consumer side, which is Applications that combine the functionality of different large models will become more and more popular.
Because the models of different modalities are trained under different data types, which means that the knowledge they learn may not overlap, so the cloud server can easily combine the functions of these large models with different modalities with different knowledge. , Combining like an assembly line, for example, you can use voice to command the car to turn on the sentry mode, and then the image recognition function detects the registered face, such as automatically opening the door when family members are in the family, and turning on the video recording function when strangers approach. This kind of highly customized application is not too difficult in the face of combined artificial intelligence, and at the same time, the user’s practicality has also been greatly improved, which will undoubtedly expand the application scope of the large model, and then allow more C-end users to use it in their daily lives. Get used to calling artificial intelligence, forming a flywheel effect. For the cloud computing platform, if it can provide different modal large models at the same time, and has this multi-modal combination capability, it can open up a new world on the C side-smart home will be more intelligent, large-scale model Internet of Things interface It will become more important, and devices that cannot access large model capabilities will be phased out. Low cost and customized functions will bring more advantages to the big cloud computing platform, so the industry may be further concentrated.
