Multiple sources have reported that Sam Altman, CEO of OpenAI, has held discussions with several major potential investors with the intention of raising billions of dollars to establish a joint venture for chip production. Meta's CEO, Mark Zuckerberg, has reiterated the company's focus on artificial intelligence (AI) and announced that their next goal is to research AGI (Artificial General Intelligence), which can achieve or surpass human-level intelligence in most fields, and has guaranteed to make this technology accessible to the public.
Image Source: Generated by Unlimited AI
Financing
OpenAI CEO Plans to Raise Billions of Dollars to Establish AI Chip Factory Network to Meet Demand
According to Sina Finance, on January 20, multiple sources revealed that Sam Altman, CEO of OpenAI, aims to raise funds to establish a factory network for semiconductor production. Altman has been working to raise billions of dollars from global investors to establish a joint venture for chip production and has held discussions with several major potential investors. Some sources indicated that Altman's fundraising efforts reflect his concerns about the lack of sufficient chips for widespread deployment as artificial intelligence (AI) becomes more prevalent. Current predictions for AI-related chip production are falling short of expected demand.
Sakana AI Completes $30 Million Seed Round Financing for Developing Generative AI Models
According to The Decoder, on January 17, AI startup Sakana AI completed a $30 million seed round financing, led by Lux Capital and Khosla Ventures, with participation from NTT Group, KDDI CVC, Sony Group, and several other investors. Headquartered in Tokyo, Sakana AI was founded by former Google AI experts Llion Jones and David Ha. The company aims to develop nature-inspired generative artificial intelligence models capable of generating text, images, code, and multimedia content.
Conversational AI Customer Service System Developer Runnr.ai Completes €1 Million Financing
As reported by IT Orange, conversational AI customer service system developer Runnr.ai announced the successful financing of €1 million, led by Arches Capital, with participation from Golden Egg Check, angel investors Roland Zeller (former GetYourGuide), Marnix van der Ploeg (former Booking.com), and hotel entrepreneur Hans Pieters. Runnr.ai is a startup that automates guest engagement through messaging, enabling hotels to communicate with guests automatically, thereby increasing engagement, saving operational time, and increasing revenue through additional sales.
Insiders: AI Startup Cohere Plans to Raise $1 Billion
According to the Financial Times, artificial intelligence startup Cohere, founded by former Google scientists, is in talks to raise up to $1 billion. Three insiders stated that the Toronto-based startup has not yet determined the valuation for the new round of financing, nor has it set a specific financing target. However, two of them indicated that discussions have included new capital of $500 million to $1 billion. Data from Crunchbase shows that this amount exceeds the total amount Cohere has raised in its four rounds of financing to date, and is expected to significantly increase the company's valuation beyond the $2.2 billion it achieved in its $270 million financing in June last year. At that time, the financing came from Nvidia, Oracle, as well as venture capital companies Index Ventures and Inovia Capital.
Mountain View Intelligence Receives Nearly 100 Million Yuan in A+ Round Financing, Financing Will Be Used to Accelerate Technological Research and Productization in the AI Field
According to "Science and Technology Innovation Board Daily" on January 22, "Mountain View Intelligence," a provider of enterprise data intelligence products and services, announced that it has received nearly 100 million yuan in A+ round financing led by Zhongke Haichuang Fund. This round of financing will further accelerate Mountain View Intelligence's technological research and productization in the field of artificial intelligence, especially in the application of large-scale model technology, which is expected to bring disruptive innovation to various industries.
Large Models
Zhipu AI Launches New Generation Base Large Model GLM-4
According to Interface News, on January 16, Zhipu AI released the new generation base large model GLM-4. According to Zhipu AI CEO Zhang Peng, the overall performance of GLM-4 has significantly improved compared to the previous generation, approaching GPT-4. Specifically, GLM-4 can support a context window length of 128k and can process text for 300 pages with a single prompt. At the same time, both text-to-image and multimodal understanding capabilities have been enhanced.
BYD Releases Vehicle Intelligence Architecture "Xuanji" and Xuanji AI Large Model
AI New Intelligence reported that on January 16, BYD announced its vehicle intelligence strategy, achieving efficient integration of electrification and intelligence through the Xuanji intelligent architecture. In addition, BYD introduced the innovative dual-cycle multimodal AI—Xuanji AI large model, applying artificial intelligence to all aspects of vehicles for the first time. The model has the largest data foundation in the industry, leading sample size, and high computing power, covering over 300 scenarios for the entire vehicle.
SenseTime's "Shusheng Puyu" 2.0 Large Language Model Open Source: 200K Context, Can Read 300,000 Chinese Characters at Once
According to IT Home on January 17, SenseTime Technology, in collaboration with the Shanghai AI Laboratory, the Chinese University of Hong Kong, and Fudan University, released the new generation large language model "Shusheng Puyu 2.0" (InternLM2). InternLM2 was trained on a corpus of 26 trillion tokens. Inheriting the settings of the first generation "Shusheng Puyu" (InternLM), InternLM2 includes two parameter specifications and bases for 7B and 20B, as well as dialogue versions, and continues to be open source, providing free commercial licenses. Currently, the data cleaning and filtering technology behind Puyu has undergone three rounds of iterative upgrades, claiming to achieve the performance of training 1T tokens using the second-generation data with only about 60% of the training data.
Lion Hunter Sky Large Model Released: 14 Billion Parameters, Can Run on a Thousand Yuan Graphics Card
According to Sina Technology on January 21, at the 2024 Fu Sheng's New Year AI Master Class and Lion Hunter Sky Enterprise Application Large Model Release Conference, Fu Sheng, Chairman and CEO of Cheetah Mobile and Chairman of Lion Hunter Sky, announced the release of the Lion Hunter Sky large model, tailored for enterprise applications, achieving the effects that can only be achieved by a hundred billion parameter large model with 14 billion parameters. It is fine-tuned for seven major application areas, such as quality education, and can run on a thousand yuan graphics card.
Stability AI Launches Stable LM 2 1.6B Model, Claimed to Outperform Microsoft Phi-2
According to VentureBeat, on January 19, Stability AI released one of its smallest models, Stable LM 2 1.6B, following the release of Stable Code 3B earlier this week, marking the second model released in 2024.
According to Stability AI, the model outperforms other small language models with fewer than 20 billion parameters in most benchmark tests, including Microsoft's Phi-2 (2.7B), TinyLlama 1.1B, and Falcon 1B. The new small Stable LM can even surpass some large models, including Stability AI's earlier Stable LM 3B model.
The First Video Large Model in China, XinYi Video Large Model, Passes Filing
AI New Intelligence reported on January 18 that XinYi Technology launched the first domestic generative artificial intelligence (large language model) focusing on the video field, XinYi Video Large Model, which has passed the "Interim Measures for the Management of Generative Artificial Intelligence Services" filing and is fully open to the public.
Tencent Launches Video Generation Model VideoCrafter2, Overcoming Data Limitations
According to Chinaz on January 19, Tencent AI Lab's team developed the video generation model VideoCrafter2, aiming to overcome data limitations and provide high-quality video diffusion models. VideoCrafter2 is a video processing tool dedicated to improving the visual quality, dynamic effects, and conceptual combinations of videos. Through unique algorithms and advanced technology, VideoCrafter2 can generate exquisite videos.
Application
Guangxi Power Grid Releases the First National Power Production Application Scenario Large Model
According to China News Guangxi on January 22, Guangxi Power Grid Company released a transmission artificial intelligence large model. Based on the "DaWatt" base of the Southern Power Grid, jointly developed by Guangxi Power Grid Company and Southern Power Grid Artificial Intelligence Technology Co., Ltd., it is the first large model for production environment applications in the domestic power industry.
According to Xiong Liang, General Manager of the Production Technology Department of Guangxi Power Grid Company, compared to traditional small models, the transmission artificial intelligence large model has superior performance in accuracy, generalization ability, and recognition efficiency. First, the parameter quantity has increased from millions to billions. Second, the defect recognition efficiency has increased by 5 times, and the accuracy has increased by 15%, enabling more precise representation of defect types and locations, solving the problem of model fragmentation, and better handling unseen power business scenario defects.
Stability AI Releases Programming Model Stable Code 3B, Can Run on Laptops
According to VentureBeat, on January 17, Stability AI released its first new AI model of 2024, the commercially licensed Stable Code 3B. This model, with 30 billion parameters, is primarily used for code autocompletion in software development and can run locally on laptops without dedicated GPUs, while also providing competitive performance and functionality compared to large models like Meta's CodeLLaMA 7B.
Stability AI claims that Stable Code 3B has demonstrated leading performance in benchmark tests across multiple languages, covering popular languages such as Python, Java, JavaScript, Go, Ruby, and C++. Early benchmark tests indicate that its completion quality can match or even exceed models twice its size.
Beijing Spring Festival Gala Launches AI Video Interactive Mini Program "Dragon Year AI Blessing You," Audience Can Obtain Their Exclusive "Metaverse" Image
According to Beijing Youth Daily on January 21, the Beijing Radio and Television Station's Spring Festival Gala launched the AI video interactive mini program "Dragon Year AI Blessing You" today, inviting the audience to obtain their exclusive "metaverse" image and become the spokesperson for the Beijing TV Spring Festival Gala. This year, the theme of the Beijing TV Spring Festival Gala is "My Spring, See Mine." Upholding the idea that this spring is blooming for every ordinary "me," this year's Beijing TV Spring Festival Gala has departed from the traditional practice of inviting celebrities as spokespersons and, using the currently popular artificial intelligence technology, has designed and developed an AI video interactive mini program to make "everyone a spokesperson for the Spring Festival Gala."
Inner Mongolia's First AI Large Model Industrial Internet Platform Released
According to People's Daily, on January 20, the Ordos Industrial Internet Developer Conference was held. At the conference, Inner Mongolia's first AI large model industrial internet platform was released. The platform, jointly created by Ordos Innovation Investment Group and Huawei, is centered on AI and MineHong, and includes a complete set of capabilities such as intelligent computing power, industrial large models, multiple preset algorithms, and application development environments.
Major Companies
Google's Two AI Scientists Plan to Resign to Establish an AI Startup
According to DoNews on January 20, according to sources, two scientists from Google's AI department, DeepMind, have been in discussions with investors to establish an AI startup in Paris. The team has discussed a round of financing that may exceed 200 million euros (about 220 million US dollars).
Insiders revealed that Laurent Sifre, a scientist at DeepMind, is in discussions with Karl Tuyls to establish the currently named "Holistic" company. They stated that the joint venture may focus on building a new artificial intelligence model. It is reported that both have submitted their resignations to the company.
Zuckerberg: Next Goal is to Build General Artificial Intelligence
According to Caixin on January 19, Meta's CEO Mark Zuckerberg reiterated the company's focus on artificial intelligence (AI). He announced that the company's next goal is to research AGI (Artificial General Intelligence) and will purchase a large number of Nvidia AI chips for this purpose. He stated that Meta is planning to build its own AGI, an AI that can achieve or surpass human-level intelligence in most fields, and has guaranteed to make this technology accessible to the public.
Opinions
Zhou Hongyi: AI Must Find Scenarios, Not Just Build Large Models
According to Sina Technology on January 21, at the 2024 Fu Sheng's New Year AI Master Class and Lion Hunter Sky Enterprise Application Large Model Release Conference, Zhou Hongyi, founder of 360 Group, and Fu Sheng, Chairman and CEO of Cheetah Mobile and Chairman of Lion Hunter Sky, engaged in a dialogue. Zhou Hongyi pointed out the need for AI to find scenarios, rather than simply aiming to build large models like OpenAI, and to consider using large models to upgrade and change existing applications. Secondly, it must be sharp and delve deeply into small scenarios, focusing on heavy verticals, rather than upgrading architectures and technology on OpenAI, as this direction is not chosen correctly and will be eliminated as OpenAI's technology advances. Thirdly, there must be a long-term perspective. AI is hot, but there is no need to be anxious. There may be some killer applications in 2024, but there may not be particularly large applications, so the focus should be on the long term.
Microsoft CEO: The Prosperity Brought by AI is No Less Than the Proliferation of PCs
According to China Entrepreneur Magazine, Microsoft CEO Satya Nadella recently participated in the 2024 World Economic Forum and was interviewed by Klaus Schwab, Chairman of the World Economic Forum and German economist. They had an in-depth discussion on the development of AI. Nadella emphasized the necessity and security of AI development, stating that the impact of this AI revolution is no less than the proliferation of PCs in the past. With the rapid development of AI, AI will provide new interfaces and models for applications. Microsoft sees the opportunity in this and believes it will bring new economic prosperity.
Research Reports
Report: In 2023, Nearly Half of the AI Large Model Products Approved for Public Access in Beijing Account for Nearly Half of the Country
According to China News, on January 21, at the 16th People's Congress of Beijing, Mayor Yin Yong stated in the government work report that in 2023, nearly half of the AI large model products approved for public access in Beijing accounted for nearly half of the country.
Report: China's Computing Power Scale Will Have a Compound Growth Rate of 20% in the Next Three Years
According to "Science and Technology Innovation Board Daily" on January 16, the "2024 Computing Power Development Trend Insight Report" released by the Ministry of Industry and Information Technology's CCID Consulting shows that in the next three years, China's computing power scale will show an explosive growth trend. It is expected that by 2026, China's computing power scale will exceed 360 EFLOPS, with a compound growth rate of 20% over three years. In addition, to meet the explosive demand for computing power brought by AI large models, the proportion of intelligent computing power and supercomputing power will continue to increase. It is expected that by 2026, the sum of China's intelligent computing power and supercomputing power will reach 40%.
Study: 98% of Executives Have Used AI to Improve Supply Chain Operations
According to Global TMT, according to the latest research from Economist Impact and DP World to be released at the World Economic Forum, business leaders remain generally optimistic about 2024, with their main driver being the belief that technology will change the efficiency and resilience of supply chains. The core of this sentiment is the widespread use of AI, with 98% of executives having used AI to innovate at least one aspect of their supply chain operations. Companies hope to further increase their adoption of technology this year.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
