Summary:: 2023 is considered to be the year of the rise of AI models, and Microsoft CEO Nadella was elected CEO of the Year 2023 for his investment and leadership in AI. Nadella has integrated AI technology into Microsoft's products and services, and has successfully commercialized AI tools such as ChatGPT. In the face of crisis, Nadella demonstrated a quick, calm, and thoughtful ability to re-emerge Microsoft as a technology innovator. "2023 is the year of AI, and we're seeing a real uptick in innovation in product manufacturing, deployment, and productivity," Nadella said. The overall score was given based on factors such as impact, public interest, novelty and importance, and Nadella received a perfect score of 100.
Summary:: The Internet police in Shangcheng District, Hangzhou recently cracked a major ransomware case, and the members of the criminal gang have relevant qualifications for network security, and use ChatGPT to carry out AI-assisted optimization of Trojan programs. **After receiving the report, a technical team was set up to investigate, and 4 suspects were successfully locked. The members of the gang had a clear division of labor in the process of committing the crime, writing the version of the ransomware, infiltrating through vulnerability scanning, planting the ransomware, and carrying out extortion. At present, the relevant criminal suspects have been taken criminal coercive measures.
Summary:In an interview, Meta's chief AI scientist Yang Likun lashed out at OpenAI, saying it had deviated from its non-profit path and become Microsoft's "contract research organization." OpenAI used a non-profit model in 2015, but in 2019 it established a subsidiary of the profit model. Yang also noted that OpenAI is no longer open and said it would not be the first company to develop artificial general intelligence (AGI). He criticized OpenAI for avoiding open source and choosing to use closed sources. The move was criticized by Yang Likun, Musk and others.
Summary:: Thunderbird has launched a large model voice assistant called Rayneo AI, which is currently in beta testing in Thunderbird's X2 AR glasses. The voice assistant provides features such as natural language conversations, trip planning, encyclopedia Q&A, and weather lookups. It is said that next year, visual understanding capabilities will be launched to support multimodal dialogue, including functions such as spatial relationships, object counting, scene understanding, and multilingual parsing. Powered by the Qualcomm Snapdragon XR2 platform, the Thunderbird X2 AR glasses have a sunglasses-like appearance, a 590mAh battery, 6GB of RAM, and 128GB of storage, as well as a binocular full-color MicroLED waveguide display with up to 1,500 nits of brightness and a 100,000:1 contrast ratio. The news has a certain influence and novelty, and it is also of high interest to the public, so the overall score is 85 points.
Summary:According to the Canalys analysis, large language models (LLMs) have driven the development of AI mobile phones. Compared with AI PCs, AI phones are closer to consumers, and manufacturers such as Apple, Huawei, Vivo, and Xiaomi have widely used on-device AI to improve imaging quality, battery life, and typing experience. According to Canalys, AI phones are a "broad and vague" concept that needs to meet a series of criteria, such as dedicated units that can accelerate AI tasks and large language models that run on the device side. The total number of smartphones is expected to reach 5 billion, of which about 5% are AI mobile phones, which will contribute to the popularization and personalization of AI applications. The functions of AI mobile phones mainly include hardware optimization, cross-device automation, image editing, social content creation, automatic answering, automatic reply, search, etc. For the success of AI mobile phones, it is important to expand the application scenarios and build AI-native applications. Smartphone manufacturers need to work closely with the developer community to gain a competitive advantage with on-device AI.
Mixtral is a sparse hybrid expert model with about 5 billion parameters, but only 13 billion models for inference. It outperformed LLAMA 2 70b in most benchmarks, with a 6x faster inference speed. It is compatible with GPT35 Matches or exceeds in most benchmarks. Available in English, French, German, Spanish, and Italian. The instruction model achieved 83 of the top scores. It performed well in coding, reaching 40 on Humaneval2% of the score.
This book provides a complete and concise overview of mathematical engineering for deep learning. In addition to providing an overview of the basics of deep learning, the book covers convolutional neural networks, recurrent neural networks, transformer models, generative adversarial networks, diffusion models, reinforcement learning, graph neural networks, and other techniques. The focus is on basic mathematical descriptions of deep learning models, algorithms, and methods. Much of the presentation of this book does not deal with computer**, neuroscience relationships, historical perspectives, and theoretical research. The benefit of such an approach is that readers with mathematical knowledge can quickly grasp the essence of modern deep learning algorithms, models, and techniques without having to look at computers**, neuroscience, or historical developments. Deep learning can be described in mathematical language and is understandable to many professionals. Readers from the fields of engineering, signal processing, statistics, physics, pure mathematics, econometrics, operations research, quantitative management, applied machine learning, or applied deep learning will be able to quickly learn about the key mathematical engineering components of the field.
For more AI tools, please refer to domestic aibard123 and github-aibard123