Abstract: 18 countries including the United States and the United Kingdom have released the first detailed international agreement on how to protect AI from unlawful acts, urging companies to build "securely designed" AI systems; Microsoft's Azure AI cloud development platform has added 40 new models such as Falcon and Stable Diffusion, covering text, images, code, and voice content generation.
Image source: Generated by Wujie AI
Hot News
Microsoft's Azure AI cloud development platform adds 40 new models such as Falcon and Stable Diffusion
According to the AIGC Open Community, Microsoft has announced the addition of 40 new models such as Falcon, Phi, Jais, Code Llama, CLIP, Whisper V3, and Stable Diffusion in the Azure AI cloud development platform, covering text, images, code, and voice content generation.
Developers can quickly integrate models into applications through APIs or SDKs, and also support tailored functions such as data fine-tuning and command optimization. In addition, developers can quickly find suitable products in the "Model Supermarket" of Azure AI through keyword search, for example, entering the word "code" will display the corresponding models.
18 countries including the United States and the United Kingdom sign an agreement urging companies to build "securely designed" AI systems
According to Reuters on November 27, the United States, the United Kingdom, and more than a dozen other countries have released the first detailed international agreement on how to protect AI from unlawful acts, urging companies to build "securely designed" AI systems.
In a 20-page document, the 18 countries unanimously believe that companies designing and using AI need to develop and deploy in ways that ensure customers and the public are protected from abuse. The agreement is not binding and mainly contains general recommendations, such as monitoring whether AI systems are being abused, protecting data from tampering, and reviewing software vendors. In addition to the United States and the United Kingdom, the 18 countries that signed the agreement also include Germany, Italy, the Czech Republic, Estonia, Poland, Australia, Chile, Israel, Nigeria, and Singapore.
Visual processing AI chip and solution company "Yinniu Microelectronics" completes a Series A financing of over 500 million RMB
According to EO News on November 27, the visual processing AI chip and solution company "Yinniu Microelectronics" announced the completion of a Series A financing of over 500 million RMB. The joint leading investors in this round are Hefei Industrial Investment and Precision Capital, with follow-up investments from Tianyu Digital Technology and some existing shareholders. The funds raised in this round will mainly be used to accelerate the research and development of next-generation chips and modules, research and development of new field product solutions, and team development and construction.
ByteDance's major retreat in games: projects to be shut down or sold, this adjustment is to focus on promoting the AIGC project
According to AI Whales, following a major retreat in education, XR, and other fields, ByteDance announced a comprehensive withdrawal from its gaming business. On November 26, ByteDance's gaming business, Daylight Studios, announced a major adjustment, shutting down all projects under development except for UGC, which has been announced internally within a week. It is rumored that for projects that have already been launched, including the well-performing "Planet Restart" and "Crystal Core," they will be operated independently and seek external buyers. This adjustment is to focus on promoting the AIGC project. Previously, ByteDance had already launched AIGC applications such as Douyin and Volcano Ark.
711 convenience stores will introduce AI on a large scale in the Japanese market from next year: used for data analysis, planning new products, etc.
According to IT Home citing Nikkei News, 711 convenience stores announced that they will introduce generative AI on a large scale in the Japanese market from next year, mainly used for data analysis, planning new products, and generating advertising materials.
The Japanese media claimed that before introducing AI, 7-11 in the Japanese market needed to go through multiple meetings and use questionnaires or consumer surveys to understand market trends and consumer preferences before launching a series of products, a process that takes about 10 months. In order to accelerate development efficiency, the company has established its own cloud-based AI model, which is said to integrate generative AI from OpenAI, Google, and Stability AI, and can collect data from various sources such as customer sales, product manufacturers, and community media, use AI to identify consumer trends and propose new product ideas, and also generate a series of advertising materials to reduce costs.
Google plans to invest $4 million in Indian startup CoRover.ai to support its developed BharatGPT
According to Webmaster's Home, Google plans to invest $4 million in the Indian startup CoRover.ai to support its developed BharatGPT. BharatGPT is an AI solution based on large language models (LLM) developed by CoRover.ai, supporting over 12 Indian languages and more than 120 foreign languages. CoRover.ai emphasizes that BharatGPT has an accuracy of 90%. Based on a vast internet database, BharatGPT has the ability to flexibly adjust and enhance content. This adaptability allows the system to integrate relevant contextual information based on factors such as region, industry, domain, customer, business, and specific use cases, ensuring that users receive current and relevant information when asking questions.
Inspur Information releases open-source trillion-level large model Source 2.0
According to Jindu Data on November 27, Inspur Information has released Source 2.0 basic large models, which include three sets of parameters: 102.6 billion, 51.8 billion, and 2.1 billion, demonstrating advanced capabilities in programming, inference, logic, and other aspects. Compared to Source 1.0, Source 2.0 focuses on improving accuracy through model structure improvements.
Good Reads
"Tian Yuandong pours cold water on OpenAI's mysterious Q* project: Synthetic data is not the savior of AGI, its capabilities are limited to simple math problems"
The discussion among experts about the Q conjecture continues. Today, AI expert Tian Yuandong publicly stated that Q* can only solve entry-level math problems, and it is highly unlikely that AGI can be achieved through synthetic data. Tian Yuandong stated:
I partially disagree with the statement that "AGI can be achieved simply by scaling up synthetic data." The reason why search is powerful is that if the environment is designed properly, it will create an infinite number of new patterns for the model to learn and adapt to. However, whether learning such new patterns requires tens of billions of data is still an unresolved issue, which may indicate that there are some fundamental flaws in our architecture/learning paradigm. In contrast, humans often find new patterns more easily through "aha" moments.
https://www.aixinzhijie.com/article/6839594
"Human race + large models, did not create the expected splash"
Compared to the enthusiasm and excitement when large models arrived, the exploration and testing of how large models are deeply applied in the industry has just begun. Finding pilot scenarios and prototype customers is the strategy of the vast majority of large model application manufacturers. However, when it comes to the end users, due to the vast differences in enterprise business logic and organizational forms, from "testing" to "comprehensive implementation," whether it's ChatGPT or "ChatPPT," there is still some confusion.
A pre-sales expert from an HR enterprise-level software said, "It's all public data, and companies dare not use it. The actual application scenarios are also very few."
https://www.aixinzhijie.com/article/6839582
"Super AI will not dominate humanity, but artificial intelligence must be open source! LeCun's latest interview attracts 3 million people to watch"
In LeCun's latest interview video, he once again revealed his views on open-source AI. "Super AI will be born one day, but it will not dominate humanity."
https://www.aixinzhijie.com/article/6839590
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。
