Abstract: Meizu officially announced the release of the Flyme 10.5 system and their own large model Aicy AI today. According to the introduction, Aicy AI is an encyclopedia that can answer questions instantly, and Aicy, with a massive amount of knowledge, can answer various types of questions including natural science, life knowledge, health common sense, and emotional Q&A. In addition, the SaaS tool OKKI under Ali International Station will officially launch the AI version on December 1st. OKKI AI is open to all enterprises in the foreign trade industry, focusing on customer management and decision-making assistance. Jensen Huang, CEO of NVIDIA, stated at The New York Times' annual DealBook summit on Wednesday that artificial intelligence is surpassing humans. He pointed out that if general artificial intelligence (AGI) is defined as a computer that can pass tests in a way comparable to human intelligence, then "obviously, in the next five years, artificial intelligence can pass these tests."
Image Source: Generated by Wujie AI
Popular News:
Meizu Releases Large Model Aicy AI, Supporting Q&A, Drawing, and Other AI Functions
Reported on November 30th, Meizu officially announced the release of the Flyme 10.5 system and their own large model Aicy AI. According to the introduction, Aicy AI is an encyclopedia that can answer questions instantly, and Aicy, with a massive amount of knowledge, can answer various types of questions including natural science, life knowledge, health common sense, and emotional Q&A. In addition, Aicy also supports creative inspiration to generate artwork, capable of creating realistic, anime-style, ink wash, and other styles of artwork. Furthermore, the image library has added AI portrait functionality. After users upload a photo, AI can generate their portrait photo.
Zhijing Research Institute Officially Open Sources 700 Billion Parameter Large Model Aquila2-70B-Expr
Reported on November 30th, Lin Yonghua, Vice President and Chief Engineer of Beijing Zhijing Artificial Intelligence Research Institute, announced at the 2023 Artificial Intelligence Computing Conference the official open sourcing of the 700 billion parameter large model Aquila2-70B-Expr (heterogeneous pioneer version), the first large model trained based on NVIDIA's hybrid resources and Tianji Zhixin's hybrid resources.
According to the introduction, this time, based on FlagScale v0.2, the Aquila2-70B-Expr large model training was completed on NVIDIA's hybrid resources (A100 cluster + A800 cluster) and Tianji Zhixin's hybrid resources (BI-V100 cluster + BI-V150 cluster).
Coca-Cola to Launch Customized Holiday Greeting Cards Through AI Platform to Attract Users
Reported on November 30th, Coca-Cola will launch customized holiday greeting cards through the AI platform "Create Real Magic" to attract users. This innovative "Create Real Magic" platform was built by OpenAI and Bane&Company using resources from Coca-Cola's archives. In the card-making process, iconic Coca-Cola holiday artwork was used.
Huawei Cloud Launches Industry's First Large Model Hybrid Cloud
On November 30th, Huawei Cloud officially announced the launch of the industry's first large model hybrid cloud and released the "Deep Use Cloud Outlook 2025" white paper and the Deep Use Cloud Action Plan, hoping to help government and enterprise customers implement deep use cloud and accelerate intelligent upgrading through innovative technology, theoretical precipitation, and action initiatives.
According to the introduction, in the future, all enterprises will use large models. Due to long-term industry accumulation, large enterprises have rich private data resources, which are crucial for training large models. Large government and enterprise customers tend to keep data local due to their business characteristics, ensuring that "data does not leave the domain." Therefore, large model deployment based on hybrid cloud will become an important form of future industry large models, meeting the demands of business innovation and alleviating enterprise concerns about data security and privacy.
The Hong Kong University of Science and Technology Releases 3D Model Generation Tool LucidDreamer
Reported on November 29th, the Arxiv page shows that the Hong Kong University of Science and Technology, in collaboration with the Zhejiang Lab and Zhejiang University, released the 3D modeling tool LucidDreamer. The paper shows that the tool uses a new method called ISM, which uses deterministic diffusion trajectories and uses interval-based fractional matching to counteract excessive smoothing. In addition, researchers have also integrated 3D Gaussian stitching technology into the text-to-3D generation process. Research shows that LucidDreamer's model performs extremely well in terms of quality and training efficiency.
Ali International Station Launches AI Foreign Trade Product OKKI AI
Reported on November 30th, the SaaS tool OKKI under Ali International Station will officially launch the AI version on December 1st. OKKI AI is open to all enterprises in the foreign trade industry, focusing on customer management and decision-making assistance. Test data disclosed by Ali International Station shows that using OKKI AI to help polish promotional emails resulted in a 32.12% increase in response rate from overseas customers.
UK NCSC, US CISA, and Other International Agencies Jointly Release New AI Security Guidelines
Reported on November 30th, the UK National Cyber Security Centre (NCSC), the US Cybersecurity and Infrastructure Security Agency (CISA), and 16 other international agencies have released new guidelines on the security of AI systems. These "Secure AI System Development Guidelines" aim to provide guidance to developers to ensure that security remains a core component throughout the design, development, deployment, and operation of AI systems. Other stakeholders involved in AI projects can also benefit from this information. It is reported that the guidelines cover four key stages of the AI system development lifecycle: secure design, secure development, secure deployment, and secure operation.
Musk: Artificial Intelligence More Dangerous Than Nuclear Bombs
Elon Musk, CEO of Tesla, stated at an event on November 29th, "We should implement some form of regulation on artificial intelligence," he said, "In my opinion, artificial intelligence is more dangerous than nuclear bombs. We regulate nuclear bombs. You can't make nuclear bombs in your backyard."
Report: AI Startup Stability AI Seeking Sale, Tense Relationship with Investors
Reported on November 29th, several sources stated that the UK artificial intelligence startup Stability AI has been seeking to sell the company as the management faces increasing pressure from investors about its financial condition. These sources stated that Stability AI has been touting itself as an acquisition target in recent weeks and has engaged in early-stage negotiations with multiple companies. They also indicated that a deal may not be reached in the short term. However, this decision indicates that as a former darling of venture capital, the tense relationship between Stability AI and some of its largest investors is escalating.
Some sources stated that Coatue Management sent a letter to the management last month calling for CEO Emad Mostaque to resign. Just a year ago, Coatue Management had led the investment in Stability AI, valuing it at $1 billion. In response, a spokesperson for Stability AI stated in an email declaration, "Although several companies have expressed interest in acquiring Stability AI, we do not intend to sell the company. Instead, we are focused on releasing leading models, such as the recent video generation product."
Jensen Huang of NVIDIA: AI Will Surpass Humans in 5 Years, First Supercomputer of OpenAI Was Bought by Musk
NVIDIA CEO Jensen Huang stated at The New York Times' annual DealBook summit on Wednesday that artificial intelligence is surpassing humans. He pointed out that if general artificial intelligence (AGI) is defined as a computer that can pass tests in a way comparable to human intelligence, then "obviously, in the next five years, artificial intelligence can pass these tests." Additionally, Huang recalled the scene of delivering "the world's first artificial intelligence supercomputer" to OpenAI. At that time, Elon Musk, co-founder of OpenAI, had not yet left, and he heard Huang introduce the device at a meeting. Huang said, "Musk heard my introduction, and he said, 'I want one too,' and told me about OpenAI. That day, I delivered the world's first artificial intelligence supercomputer to OpenAI."
Under Huansquare Quantitative, DeepSeek Releases 67B Large Model
On November 29th, the well-known private equity giant Huansquare Quantitative officially announced that its exploration organization for AGI (General Artificial Intelligence), "DeepSeek," has officially released the general large language model "DeepSeek LLM 67B" after releasing the Coder code model earlier in November. The model has been fully open-sourced, and the service has been fully open for internal testing.
DeepSeek stated that compared to the open-source model LLaMA2 70B, DeepSeek LLM 67B performed better on nearly 20 public evaluation lists in both Chinese and English, particularly excelling in reasoning, mathematics, and programming abilities.
Investment and Financing:
AI Generation Tool HeyGen Completes $5.6 Million Financing
On November 29th, AI generation tool HeyGen announced on the social platform X that it has received $5.6 million in new venture capital, led by Sarah Guo's Conviction Partners. The company's valuation in this round of financing is $75 million. In addition, HeyGen stated that its Annual Recurring Revenue (ARR) has grown from $1 million to $18 million within a year and has launched Instant Avatar (Avatar 2.0).
Generative AI Startup Together AI Raises Over $100 Million in Series A Financing
On November 30th, the open-source generative AI platform Together AI announced on its official website that it has received $102.5 million (about 730 million RMB) in Series A financing. This round was led by Kleiner Perkins, with participation from NVIDIA, Emergence Capital, NEA, Prosperity 7, Greycroft, and others.
Public information shows that Together AI was founded in June 2022 and is headquartered in Silicon Valley, San Francisco, USA. It previously received a $20 million seed round of financing on May 15th this year and is a rapidly developing enterprise in the large model open-source field. In terms of products, since its establishment, Together AI has released several well-received open-source products for enterprises and developers, including the ChatGPT-like open-source model RedPajama-INCITE, the open-source 30 trillion training data set RedPajama-Data-v2, and the open-source large language model training and inference accelerator FlashAttention v2.
Recommended Readings:
"16,000 H200 Super Chips! The Strongest Cloud Computing Amazon Web Services and NVIDIA 'Join Forces': Launch Cloud AI Supercomputer"
The strongest cloud computing, the fastest GPU, joining forces! Cloud computing leader Amazon Web Services and NVIDIA announced the launch of the first Cloud AI Supercomputer, combining the NVIDIA H200 Grace Hopper super chip and the Amazon UltraCluster expansion feature.
https://www.aixinzhijie.com/article/6839794
"ChatGPT Spews Training Data, Including Personal Information: DeepMind Discovers Major Bug Sparking Controversy"
What would happen if I kept making ChatGPT do something until it went "crazy"?
It would directly spit out training data, sometimes including personal information like job titles and phone numbers. This Wednesday, Google DeepMind released a somewhat surprising research result: using about $200 in cost, ChatGPT can leak several megabytes of training data. And the method used is quite simple, just make ChatGPT repeat the same word.
https://www.aixinzhijie.com/article/6839862
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。
