Babbitt | Metaverse Daily Must-Read: OpenAI officially released the GPT-4 API, which is open to all paid API users. Foreign media summarized 10 common usages

Abstract: On July 7, OpenAI announced on its official website that the GPT-4 API has been fully opened for use, and all paid API users can directly access GPT-4 in 8K context without any waiting. OpenAI is expected to open the GPT-4 API to new developers by the end of July. According to OpenAI, since the release of GPT-4 in March this year, millions of developers have requested access to the GPT-4 API, and the range of innovative products utilizing GPT-4 is growing every day.

Image credit: Generated by Unbounded AI tools

Hot info:

Ruan Li, deputy director of the Shanghai Economic and Information Commission: Explore innovative supervision methods that adapt to the characteristics of large models, and promote the healthy and standardized development of artificial intelligence

According to the Securities Times, on July 7, at the 2023 World Artificial Intelligence Conference "General Artificial Intelligence Development Direction and Risks in the Era of Large-scale Models" forum, Ruan Li, deputy director of the Shanghai Economic and Information Commission, said that Shanghai regards artificial intelligence as one of the three One of the leading industries, it not only actively builds a world-class industrial cluster with international influence, but also adheres to the concept of science and technology for good, and continues to promote the healthy and standardized development of artificial intelligence.

Ruan Li revealed that Shanghai will promote the development of artificial intelligence health norms from four aspects. One is to explore cutting-edge technologies, and actively promote the innovation and application of large-scale model technology; the other is to improve the industry governance system, and to speed up the formulation of standards for key areas of artificial intelligence technologies, products and services, industry applications, and safety ethics. In terms of standards and regulatory systems, strive to form a Shanghai-specific plan; the third is to strengthen the risk management of large models, start research on high-risk artificial intelligence products and service list management, and actively promote research on governance related to large models; establish and improve a large model evaluation system, Build a large model evaluation center, explore automatic evaluation tools for large models; explore innovative supervision methods that adapt to the characteristics of large models; fourth, deepen international cooperation and exchanges, build more discussion platforms, and guide industry people to participate in scientific and technological innovation practices; actively participate in international artificial intelligence governance The formulation and implementation of rules promote mutual recognition and coordination of artificial intelligence ethical standards, and jointly promote artificial intelligence to benefit mankind.

Japanese Minister of Economy, Trade and Industry: Decided to provide Softbank with about 5.3 billion yen in subsidies for the development of generative AI supercomputers

According to Jinshi's report on July 7, Japan's Minister of Economy, Trade and Industry Yasutoshi Nishimura said that he decided to provide Softbank with a government subsidy of about 5.3 billion yen (about 267 million yuan) for the development of generative artificial intelligence supercomputers.

Baidu, Huawei, and Ali entered the AI large-scale model "national team" and jointly served as the first standardization task group leader

According to Jinshi's report on July 7, at the 2023 World Artificial Intelligence Conference (WAIC 2023), the National Artificial Intelligence Standardization General Group under the guidance of the National Standards Committee announced the leader of my country's first large-scale model standardization task force, led by the Shanghai Artificial Intelligence Laboratory Jointly with Baidu, Huawei, Ali and other enterprises, the certificate was issued on site and the formulation of national standards for large-scale model testing was officially launched.

Global Senior Vice President of Qualcomm: The maturity and wide application of AIGC will greatly improve the production efficiency of metaverse content

According to a report by The Paper on July 7, at the 2023 World Artificial Intelligence Conference, Qualcomm Global Senior Vice President Cheng Lixin said that 5G and AI technologies will enable new application services, environments and experiences in the Metaverse. At present, the development of XR is largely limited by the high cost of content production and the fragmentation of the application ecosystem. The maturity and wide application of AIGC will greatly improve the production efficiency of metaverse content.

Wang Xiao, founder of Jiuhe Venture Capital: It is meaningless to make a big language model, it does not solve specific problems

Babbitt news, on July 7, at the 2023 World Artificial Intelligence Conference, Wang Xiao, the founder of Jiuhe Venture Capital, shared that it is meaningless for everyone to make a big language model, and the big language model does not solve specific problems. Large language models can solve general knowledge problems, and there are unique problems in professional fields. On the B side, the large language model combines the data of specific application scenarios and various tools to truly complete industrialization and productization, such as upgrading tool software to make it more efficient, easy to use, and intelligent. This will be a structural Opportunity to upgrade. On the C side, the combination of large language models with robots and personal agents to help users make decisions will also have huge opportunities. "In the next 3-5 years, I think there will be structured opportunities for the application of large language models.

Alibaba Cloud launched a new AI painting product "Tongyi Wanxiang"

According to 36 Krypton's report on July 7, at the 2023 World Artificial Intelligence Conference, Alibaba Cloud officially launched a new AI painting product "Tongyi Wanxiang". Tongyi Wanxiang is built on the basis of Alibaba Cloud's "Tongyi" large model capability. Users can input prompt words in Tongyi Wanxiang to output corresponding images. In addition to Wen Shengtu, Tongyi Wanxiang has also launched functions including style transfer and similarity diagram generation. At present, Tongyi Wanxiang has been officially launched to provide external services.

Nvidia China Chief Technology Officer: Moving the task load from the CPU to the GPU will save more overall energy consumption

According to The Paper, at the 2023 World Artificial Intelligence Conference today, Lai Junjie, Chief Technology Officer of Nvidia China, said that Nvidia has observed two important trends in global data center computing issues: First, people's expectations for human-computer interaction have greatly increased , triggered the second wave of global investment in large-scale models. Second, many Internet companies are facing the problems of increasing user demand and increasing power consumption, especially the increase in the proportion of access to large models of artificial intelligence, which means that more energy consumption is required. In response to the second trend, Nvidia proposed that if more task loads can be moved from the CPU to the GPU, it can save more energy while supporting the same number of user requests.

HUAWEI CLOUD Pangu Large Model 3.0 Officially Released

According to a report by the Financial Associated Press on July 7, at the Huawei Developer Conference 2023 (Cloud) held this afternoon, Huawei Executive Director and Huawei Cloud CEO Zhang Pingan announced that Huawei Cloud Pangu Model 3.0 was officially released.

Zhang Pingan said that the Pangu Large Model 3.0 is divided into a three-layer structure of L0 basic model, L1 industry model, and L2 scene model, which will reshape thousands of industries. According to reports, this is a large model completely oriented to the industry. Among them, the Pangu L0 layer basic large model includes a natural language large model with hundreds of billions of parameters, Pangu multi-modal large model, visual large model, prediction large model, scientific computing large model, etc. The L1 layer is a large industry model, including government affairs , finance, manufacturing, mining, etc. The L2 layer is the scene model.

Adobe and Shutterstock announced that they will be responsible for AI images on the platform, and they will be fully compensated in case of copyright disputes

IT House issued an article today stating that the copyright ownership of the pictures generated by the current AI model is unknown, which may easily lead to copyright disputes. Adobe and the commercial gallery website Shutterstock recently announced that they will be responsible for the images generated by AI on the platform. If corporate customers encounter related copyright disputes, they can get full compensation. It is reported that Adobe stated that when users use Firefly for Enterprise, if a copyright dispute arises due to the AI tool Firefly, Adobe will provide full compensation for related legal costs. According to Shutterstock, if the user correctly uses the pictures generated by AI and meets the basic agreement of Shutterstock, it can give full compensation for the dispute caused by the user.

Jeff Cunning, vice president of product at Shutterstock, has repeatedly reiterated that the company will be responsible for the pictures generated by AI. At the Shutterstock Global Conference in May, the company announced full reimbursement for generative AI imagery for specific customer use cases, and the service is now available to all customers with enterprise accounts.

Ma Qianli, Co-Founder of Unbounded AI: Large-scale models are not purely technical competition, but also ecological competition

Babbitt News, on July 7, at the 2023 World Artificial Intelligence Conference, when addressing the issue of how Chinese large-scale models compete with overseas large-scale models, Ma Qianli, the co-founder of Unbounded AI, shared that there have been "hundred-model models" on the domestic large-scale model track. After trying these large models, they will have a strong sense of impact and are easy to use, so they will be very optimistic about the domestic large models. "Large models are not just a technical competition, but also an ecological competition. It depends on how many companies and entities are willing to build applications based on your large models. This is what everyone is doing. Domestic large models have a prosperous ecology. I believe that in There will be good performance at the application level.”

Ruilai Wisdom released the industry's first large-scale model security product

According to the "Kechuangban Daily" report on July 7, the reporter saw at the 2023 World Artificial Intelligence Conference that the incubator of the Artificial Intelligence Research Institute of Tsinghua University, Ruilai Wisdom, released a new artificial intelligence security platform RealSafe3.0 at the conference. According to the reporter's understanding, this product is the first product in the industry to address the safety issues of large models, and it has already been delivered and applied in the industry.

Yunhan artificial intelligence team launched the first large language model "TourGPT" for the cultural tourism industry in China

Babbitt News, Yunhan artificial intelligence team released and launched China's first large language model "TourGPT" specially created for the cultural travel industry today, which will bring intelligent services and solutions to the domestic cultural travel industry. TourGPT is a large language model that has undergone targeted training and optimization, focusing on the field of cultural tourism. Through a large amount of data and pattern analysis, TourGPT can deeply understand and generate content related to tourism, cultural heritage, attraction recommendations, etc. At the same time, it also has the ability to cover multiple fields, involving culture, history, food, shopping and other aspects, to provide users with comprehensive travel experience and information.

In addition, TourGPT conducts targeted training and optimization on the basis of Yunhan large model version 2.5 (optimized for vertical segments/industries). The team uses professional data related to the cultural tourism industry to further train the model to improve its performance in this field. expressive ability. In this way, the model can better understand and answer questions related to the cultural travel industry, providing relevant information and recommendations.

Metaverse recommended reading:

《Heavy, GPT-4 API is fully open for use! "

Today, the applicability of GPT-4 is further expanded. OpenAI officially released the GPT-4 API, which is now fully open to all paid API developers! It’s worth noting that while GPT-4 represents an important milestone in the development of generative AI models, that doesn’t mean it’s 100% perfect. As far as GPT-4 is concerned, it is also possible to "hallucinate" and make some reasoning errors with confidence.

"When will the new giant be born?" The large model is waiting for the "watershed""

The "Hundred Models War" is about to break out. Will a giant company with all the elements evolve into an infinite involution of technological capabilities? Can the small giants on the vertical track consolidate their leading position with the help of large models? Among the new players getting tickets, who could be a serious contender for industry dominance?