OpenAI
It is reported that OpenAI will take additional measures from July 9 to block API traffic from "non-supported countries and regions", while OpenAI announced that there is no Chinese mainland and Hong Kong and Macau in the list of countries and regions that support access to API services. The company's official website also shows that in addition to the countries and regions on the list, accessing or providing access to OpenAI's services may result in account blocking or suspension.
The following is a comprehensive report brought to you by "Phoenix Art".
01
“OpenAI停服”
"The thunder is loud, the rain is small", some AI entrepreneurs described OpenAI as saying that it would block domestic API (application programming interface) calls from July 9 this year.
In the eyes of some people, although "more than half of the top large-scale model products in China in 2023 are 'shell' ChatGPT (that is, calling open-source large-scale model APIs to provide services, rather than self-developed models)", OpenAI's interruption of ChatGPT's call did not usher in the imaginary industry earthquake.
It is understood that there are currently two mainstream channels for using OpenAI's technology in China, one is to connect with the API officially provided by OpenAI; The second is to connect with the OpenAI technology provided by Microsoft's intelligent cloud Azure, and Microsoft has not yet sent a similar API ban to developers.
The analysis points out that at present, many startups in China are connected to the OpenAI API to make large models or AI applications, which may be directly affected. Some market analysts believe that OpenAI's closure of the API actually means that it has ceded the Chinese market, and this incident will promote the increase in the market share of domestic large model companies in the short term. At the same time, it will also force Chinese large-scale model companies to accelerate independent research and development, and urge more start-ups to choose domestic large-scale models to avoid possible risks in the future and reduce dependence on external technologies.
It is generally believed in the industry that China and the United States currently have their own advantages in the field of artificial intelligence. From the perspective of China, up to now, China has launched about 305 large models, of which about 140 large models have completed the filing of generative AI services, accounting for about 45.9% of the total number of 305 large models. Although the United States is leading in the development of basic models, China occupies a dominant position in the number of global AI patents and the number of industrial robot installations.
China accounted for 61.1% of the world's AI patents in 2022, far exceeding the 20.9% of the United States. In 2022, the installation of industrial robots in China will be 290,300 units, 7.4 times that of 39,500 units in the United States. However, INFORMATION reports that Open AI earns up to $1 billion in annual revenue from model access.
In the era of large models, computing power is the foundation. Major technology companies around the world are increasing capital investment in computing power, and many local policies related to computing power construction have been released one after another, bringing development opportunities to the industrial chain. In Beijing, we will continue to promote the construction of computing power infrastructure and build a Beijing-Tianjin-Hebei-Mongolia computing power supply corridor with Inner Mongolia-Hebei-Beijing-Tianjin as the main axis. In Shanghai, artificial intelligence is also one of the three leading industries that Shanghai is focusing on, and will focus on model algorithms, intelligent computing power, embodied intelligence and other tracks, so as to realize the precise docking of artificial intelligence technology products and application needs, and accelerate the generation of new quality productivity. Shenzhen issued a plan to build a city-level intelligent computing platform.
Industry analysts pointed out that on the C-side, although ordinary consumers' willingness to pay for AI products has risen, it is difficult for revenue to cover the high cost of large model training and operation. In addition, many technology companies tend to focus too much on AI technology itself, and ignore the development of the consumer market and the exploration of consumer demand. On the enterprise side, many traditional enterprises need to consider ROI and data security when adopting AI technology, and in addition, integrating AI into previous workflows and subsequent maintenance will incur high costs, which makes the technology and user needs mismatched.
In the past two years, generative AI has played an increasingly important role in the global economy. According to a recent Deloitte report, the adoption of generative AI in developing economies is 30% higher than in developed economies. In China, China has a large market scale and rich application scenarios, which provides a broad space and conditions for the landing and application of large models. Deloitte China Deputy CEO Liu Minghua believes that in order to narrow the gap in global competition, China should strengthen the underlying technology.
Market participants pointed out that China's large-scale market makes China slightly inferior to the United States in terms of underlying R&D technology, but still has the opportunity to catch up with the United States. However, in the process of catching up, more and more obstacles will be encountered, which may be a definite trend and a challenge for Chinese related enterprises. The United States previously announced a draft investment ban aimed at restricting U.S. entities from investing in China in three high-tech fields: semiconductors and microelectronics, quantum information technology, and artificial intelligence. China's Ministry of Commerce expressed strong dissatisfaction with this and called on the US side to stop politicizing and weaponizing economic and trade issues. At the recent Global Digital Economy Conference, Orens, chairman of the National Committee on U.S.-China Relations, also believes that China and the United States should strengthen cooperation in the field of artificial intelligence and digitalization.
02
"Who can be a substitute?"
Undoubtedly, OpenAI was once the best choice for domestic AI entrepreneurs. Now, becoming OpenAI's replacement is becoming the next scenario for domestic large models to compete for.
In fact, domestic large models have begun to encroach on OpenAI's share of the market to a certain extent. The temptation brought by the huge price difference undoubtedly makes AI entrepreneurs "really fragrant". "The same 1 million Tokens (the smallest processing unit of large model text) can be used by 100 users for 5 months, and it costs 100 yuan to buy GPT 4.0, but the domestic large model DeepSeek only needs one yuan." Fu Han told the alphabet list that now under the price war of the head large model manufacturers, domestic AI entrepreneurs use domestic large models, and the cost of Tokens in a single month may be hundreds of times different from that of calling GPT 4.0.
With the news that OpenAI will cut off supply, major manufacturers have dispatched and begun to "cut off" developers. Up to now, including Zhipu AI, Zero One Everything, Dark Side of the Moon, Baichuan Intelligence, SenseTime, Baidu, Alibaba, Tencent, etc., have launched OpenAI's zero-cost migration plan. At the same time, some large-scale model manufacturers also intimately give away 10 million or even 100 million Tokens.
▲Stanford University's AI Town Project
Previously, "prepare individual licenses, service account authentication, domain names, rent a domestic server, a foreign server, and about 2,000 yuan can make a small application of ChatGPT." "In 2023, while the domestic large model is catching up with ChatGPT and flawlessly launching AI applications, many people will borrow the ChatGPT port, and 2-3 people will be able to earn more than 10,000 small workshops a day, making a lot of money.
As the head application of AI search, Perplexity, launched in December 2022, also once relied on OpenAI's GPT 3.5 model, in this regard, Perplexity CEO Srinivas bluntly said, "A shell product with 100,000 users is more meaningful than having its own model but no users." ”
However, as large model manufacturers begin to launch APP applications, which can even be directly free, and the functional scenarios are more abundant, the life of many small workshops is becoming more and more difficult. Whether it is accessing OpenAI through Microsoft Cloud or switching to a domestic large model, it is a definite fact that the traffic dividend brought by OpenAI's popularity in 2023 is disappearing for AI entrepreneurs with large models in the shell.
In just a year and a half, OpenAI's API dividends have faded. With the improvement of the underlying capabilities of domestic large models, leading manufacturers have squeezed into the AI application layer to compete for the remaining traffic. It is becoming less and less possible to try to make a "big hit" based on the fine-tuning of a large open source model.
03
"Is AI art affected in China?"
Earlier this year, OpenAI announced the launch of a new generative AI model, "Sora". With text commands, Sora can output up to 60 seconds of video directly and include highly detailed backgrounds, complex multi-angle shots, and emotionally charged characters. Following the image generation system DALL· After E and the chatbot ChatGPT, OpenAI expanded its advanced AI technology to the video field, becoming an important milestone in the realization of AGI (Artificial General Intelligence).
OpenAI made it clear on Sora's official blog that Sora's model is not perfect at present, and it is still in the early stage of world model research and application. OpenAI has also invited a group of visual artists, designers, and filmmakers to join in the hope of receiving valuable feedback to advance the model and better empower creatives.
In recent years, with the advent of AI painting tools such as Google Brain's Imagen, OpenAI's DALL-E, and midjourney, the gap between conceptual vision and artistic presentation has been increasingly bridged, and "everyone is an artist" seems to be turning from a slogan into a reality. Because scientists have invented a computer language model that generates images directly from text, as long as we enter the prompt words we set, the corresponding image can be generated by artificial intelligence. These AI painting tools transcend the barriers between "words" and "images" through the magic of neural network examples, big data models, and style transfer, and the quality of the works produced is getting higher and higher
Although AI art has a wide range of applications on an international scale, there are still challenges in specific vertical applications such as the digital production of Chinese culture and art. There is no common standard or benchmark, which can lead to fragmentation of research and application. Just as ImageNet is driving progress in the field of image recognition, there is an urgent need for the AIGC field to establish corresponding standards to lead and integrate the development of this field. The critique of generative art is complicated by the depth of expertise and subjective aesthetics it involves, and there is an urgent need for expert input from the art world (e.g., RLHF) to establish evaluation criteria. The co-developed evaluation model (Reward Model) can provide expert-level scoring for automatically generated artworks independently of the generative model, aiming to provide a more accurate and scientific evaluation mechanism for the generative field and strengthen the connection between artists and technical researchers.
However, the "shutdown" of OpenAI will prevent artists and designers who rely on OpenAI's technology from accessing these important generative tools, affecting their efficiency and innovation in digital art, conceptual design, and text creation.
The application of AI in the field of art involves the deep integration of technology and culture, and the comprehensive strength of the two is reflected behind it. The suspension of services this time undoubtedly makes people rethink the future development direction of Chinese AI art. At the same time, the application of AI to artistic creation can promote the improvement of AI algorithms. By studying and analyzing the performance of algorithms in different application scenarios, people can continuously reflect on the limits, principles and future development of AI algorithms. Artistic creation can not only expand AI application scenarios, but also provide experimental data and practical basis for the improvement of AI algorithms. Cross-border cooperation can simultaneously promote the development of technology and the digital transformation of culture and art, thereby promoting the process of Chinese modernization.
Therefore, the deep integration of art and AI will become an important direction of cultural digitalization, which is also the important significance of studying AI in the national cultural digitalization construction. The suspension of OpenAI's service may also force China's art system to more urgently establish a more localized and solid generative logic and soil.
At this time, the ban of OpenAI's API in the domestic market is quite a meaning of "one whale falls and all things are born". Domestic large-scale model manufacturers have launched their own migration solutions to attract users through discounts, free services and technical support. Large-scale model service providers are once again pulled to the starting line to race, and whoever can use the most attractive scheme to reassure users will have a greater chance of standing out and becoming the leader in the second half.
Below, "Phoenix Art" also presents you with several more active domestic models. In your opinion, who can be a real "replacement"?
Zero One Thousand Things - "Yi API Two-fold Replacement Plan"
Zero One has launched the "Yi API Two-Fold Replacement Plan", which has launched a smooth migration service for OpenAI users to the Yi series of large models, helping potential users to calculate the large model in detail.
Zhipu AI - "Special Moving Plan"
Zhipu provides a special move plan for OpenAI API users to help them switch to the GLM series of models. Zhipu provides a series of services to attract users who need to migrate:
Baichuan Intelligence - "Migration only takes five minutes"
Baidu Intelligent Cloud - the cloud of the hometown, the main free
Baidu Intelligent Cloud Qianfan launched a large-scale inclusive plan, and withdrew a number of free product calls and services for newly registered enterprise users:
Alibaba Cloud - Speak with strength and price
Alibaba Cloud also invites users to migrate to the Tongyi API. And succinctly and clearly threw out the reason:
Some of the pictures in this article are from the official account of Phoenix Satellite TV and the Internet
(Phoenix Art Comprehensive Report Editor/dbk Ning Editor-in-charge/DBK)
凤凰艺术
最具影响力的全球艺术对话平台
艺术|展览|对话
Copyright Notice: All works marked "Source: Phoenix Art" on this website are works that this website legally owns the copyright or has the right to use, if you need to obtain cooperation authorization, please contact :[email protected], [email protected]. If the work is authorized to be used by this website, it should be used within the scope of authorization, and indicate "Source: Phoenix Art". The above works may not be reproduced, excerpted or used in other ways without the authorization of this website.