Recently, Musk officially launched the brand new chatbot Grok 3 in a live stream and disclosed the astonishing cost of training the model. Grok 3 is currently available to Premium + subscribers and has shown outstanding performance in various evaluation areas, even surpassing other competitors in the market, such as Gemini, DeepSeek, and ChatGPT. During the live stream, Musk mentioned that the training process for Grok 3 consumed a total of 200,000 NVIDIA GPUs, a figure that is truly shocking.

In today's era, artificial intelligence (AI) is rapidly integrating into various industries, and the education sector is undoubtedly one of its core application scenarios. The rise of AI technology has provided new opportunities for reducing costs and increasing efficiency in education, making high-quality, inclusive, and personalized education possible. Recently, multiple education companies have adopted the DeepSeek large model, indicating that the commercialization process of AI education products is accelerating. According to a research report by Zhongxin Jiantou, AI technology can deeply empower the education industry by enhancing delivery efficiency and personalizing experiences.

The ByteDance Doubao large model team announced today the successful development of a new sparse model architecture called UltraMem. This architecture effectively addresses the high memory access issues during the inference of MoE (Mixture of Experts) models, improving inference speed by 2 to 6 times compared to MoE, and reducing inference costs by up to 83%. This groundbreaking advancement opens a new path for efficient inference of large models. The UltraMem architecture successfully resolves the memory bottleneck during inference of MoE architectures while maintaining model performance. Experimental results show that the parameters and activation conditions are the same.

On February 11, 2025, at the World Governments Summit held in Dubai, UAE, Baidu founder Li Yanhong engaged in an in-depth dialogue with UAE AI Minister Omar Sultan Al Olama to explore future development trends in AI infrastructure. During the discussion, Al Olama mentioned the recent global attention towards DeepSeek, which caused significant fluctuations in the stock prices of several major chip manufacturers and exchanges, and asked Li Yanhong for his views on the future of data centers and AI infrastructure.

Recently, the World Government Summit (WGS) grandly opened in Dubai, UAE, attracting attention from elites around the globe. During the summit, Baidu founder Li Yanhong engaged in an in-depth dialogue with the UAE's AI minister to explore the latest development trends in the AI field. In the conversation, Li Yanhong specifically mentioned Baidu's innovative achievements in AI inference—DeepSeek. He stated that innovation is not limited to the AI field or the IT industry, but is widespread across various sectors, with the core of most innovations closely related to cost reduction.

Recently, Amazon CEO Andy Jassy discussed the development of artificial intelligence (AI) technology and its cost issues during a conference call with analysts. He stated that with the continuous emergence of new technologies, particularly the introduction of models like DeepSeek, the costs of AI inference will decline significantly. This change will enable more businesses to easily apply AI to their operations, especially in the fields of inference and generative AI. Image source note: Image generated by AI, licensed by Midjourney.

Recently, AI researchers from Stanford University and the University of Washington successfully trained an AI reasoning model named S1, with a training cost of less than $50 and very low cloud computing credits required. This research was published last Friday, indicating that S1 performs comparably to OpenAI's O1 model and DeepSeek's R1 model in mathematical and programming ability tests. The code and data for S1 have been made publicly available on GitHub for other researchers to use. The research team reported that they drew from a

Chinese artificial intelligence company DeepSeek recently released the open version of its inference model R1, which has quickly sparked discussions in the tech community. Its remarkable breakthrough achievements left venture capitalist Marc Andreessen in awe, describing it as 'one of the most astounding and impressive breakthroughs I've ever seen,' and it has demonstrated capabilities in AI benchmarking that rival or even surpass those of OpenAI's o1 model. Notably, DeepSeek claims that its model training cost is only $5.6 million, whereas leading U.S. companies require hundreds of millions.

Against the backdrop of rising labor costs, the UK retail industry is rapidly moving towards an era of automation. From electronic shelf labels to self-checkout machines, many new technologies are gradually transforming the familiar supermarket shopping experience. With the national minimum wage increase and higher employer national insurance contributions, retailers are significantly increasing their investments in automation technology, hoping to address the rising labor costs by improving production efficiency. Recent holiday trading updates show that investment in automation has become a focal point for major retailers. A survey indicates that nearly one-third of retail financial leaders...

Recently, the DeepSeek team in China has launched its latest open-source large model R1, which has gained widespread attention. The performance of the R1 model is exceptionally outstanding, surpassing OpenAI's o1 model in multiple tests, especially excelling in assessments related to mathematics and programming. In the latest AIME2024 test in the United States, R1 scored 79.8, outperforming o1's score of 79.2. In the MATH-500 test, R1 achieved a score of 97.3, again leading over o1.

In today's rapid development of Large Language Models (LLM), the training and inference costs of models have increasingly become a focus of research and application. Recently, the Tencent Hunyuan team released an important study that delves into the 'Scaling Laws' of low-bit floating point quantization training, which refers to the principles governing the scale of floating point quantization training. The core of this research lies in exploring how to significantly reduce computational and storage costs without sacrificing performance by lowering the precision of the model.

According to Intelligent Eruption, embodied intelligent agent operating system developer Yunjinwei recently announced the completion of an angel round financing of several million yuan, led by iFlytek. This round of financing will be used for core technology product development and market ecosystem construction. Founded in June 2021, Yunjinwei focuses on the development of embodied intelligent agent operating systems. The company was founded by Dr. Wang Wenyu, former general manager of the hardware division of Yitu Technology, and co-founder Dr. Zhou Chang was formerly responsible for visual large model development at Alibaba's DAMO Academy for the Urban Brain. The team has received support from Ningbo's Yaojiang Talent Plan and Yongjiang Talent Engineering Plan.

As the potential of artificial intelligence in healthcare draws increasing attention, Brazilian startup Carecode is emerging in this trend. Carecode focuses on leveraging AI agents to handle various tasks before and after medical appointments, aiming to lower healthcare costs and improve outcomes. CEO Thomaz Srougi stated in an interview that the importance of not only face-to-face meetings with doctors cannot be overlooked, but also the stages before and after appointments are crucial. His medical background stems from a decade of industry experience, during which he founded.