Welcome to the 【AI Daily】 column! Here is your guide to exploring the world of artificial intelligence every day. We present you with the hottest topics in the AI field, focusing on developers to help you gain insights into technology trends and understand innovative AI product applications.
Fresh AI products Click to learn more: https://top.aibase.com/
1. Apple and Alibaba to develop AI features for iPhone in China
Apple has partnered with Alibaba to jointly develop AI features for iPhone users in China. This move marks Apple's further expansion into the AI field in the Chinese market. After evaluating several potential partners, Apple ultimately chose Alibaba due to its technological advantages and capabilities in AI. This collaboration not only demonstrates Apple's commitment to the Chinese market but also aims to provide users with a smarter and more convenient experience.
【AiBase Highlights:】
🤝 Apple collaborates with Alibaba to develop AI features for iPhone users in China.
🔍 During the evaluation process, Apple found shortcomings in customer support from teams like Deepseek.
📈 This partnership reflects Apple's focus on the Chinese market and Alibaba's strong capabilities in AI.
2. Reports indicate Baidu will release next-generation AI model Ernie 5.0 this year
According to media reports, Baidu plans to launch its next-generation AI model, Ernie 5.0, this year. This model is referred to as a "foundation model" and is expected to significantly enhance multimodal capabilities, although specific functions have not yet been disclosed. This news coincides with Apple shifting its potential customers to Alibaba, leading to market speculation that Baidu's move aims to respond to market changes and stabilize its stock price and market position.
【AiBase Highlights:】
🚀 Ernie 5.0 will significantly enhance multimodal capabilities, expanding the applications of AI.
📉 Baidu's move is seen as a response to market changes, aiming to stabilize its stock price and market position.
🛠️ Specific functions have yet to be disclosed, so further details are awaited in future releases.
3. National Supercomputing Platform launches AI Partner Program: DeepSeek API free for three months
The National Supercomputing Internet Platform has launched an AI Ecosystem Partner Acceleration Program aimed at providing convenient AI technology application support for enterprise users. This program includes three months of free access to the DeepSeek API interface, along with a computational resource pool of ten million core hours. This initiative not only showcases the National Supercomputing Platform's proactive efforts in promoting the popularization of AI technology but also provides strong support for enterprises to lower the barriers to AI application and accelerate digital transformation.
【AiBase Highlights:】
🆓 Offers three months of free access to the DeepSeek API interface, supported by a ten million core hour computational resource pool.
🌐 The platform has achieved interconnectivity with supercomputing and intelligent computing centers across 14 provinces, aggregating 6,500 computational products.
⚙️ Supports multiple versions of the DeepSeek-R1 model, allowing flexible deployment of private API services and AI web applications.
4. ByteDance's Doubao UltraMem architecture reduces large model inference costs by 83%
ByteDance's Doubao team has announced the development of the UltraMem architecture, which successfully addresses the high memory access issues of the MoE model during inference, significantly improving inference speed and reducing costs. This innovation provides a new solution for efficient inference of large models, demonstrating excellent performance and cost-effectiveness, especially in large-scale applications, marking a significant advancement in inference technology.
【AiBase Highlights:】
⚡ The UltraMem architecture achieves 2-6 times faster inference speed and reduces inference costs by up to 83%.
🧠 Under the same parameters and activation conditions, the UltraMem model outperforms MoE, with memory access costs close to those of dense models.
📈 This architecture lays the foundation for building models on the scale of billions, helping to promote the large-scale application of models.
5. Xiaohongshu's open-source speech recognition model FireRedASR excels in Chinese recognition accuracy
The FireRed team from Xiaohongshu has launched a new open-source speech recognition model, FireRedASR, marking a significant breakthrough in Chinese speech recognition technology. This model performs excellently across multiple standard test sets, achieving a character error rate (CER) of 3.05%, an 8.4% reduction from the previous best model. FireRedASR features two core structures, focusing on speech recognition accuracy and inference efficiency, suitable for various language environments.
【AiBase Highlights:】
🎤 FireRedASR is a newly released open-source speech recognition model from the Xiaohongshu team, excelling in Chinese recognition accuracy.
🚀 The model is divided into FireRedASR-LLM and FireRedASR-AED, targeting accuracy and efficiency needs respectively.
🌍 FireRedASR performs excellently in various scenarios, suitable for Mandarin, Chinese dialects, and English.
Details link: https://huggingface.co/FireRedTeam
6. ChatGPT energy consumption revealed: only 0.3 watts per response, much lower than rumored!
Recent research has revealed the energy consumption of ChatGPT, showing that it is significantly lower than earlier estimates. According to Epoch AI's analysis, using GPT-4o consumes only about 0.3 watt-hours per query, a figure lower than many household appliances. Despite public concerns about AI energy consumption, the current situation has not been accurately reflected.
【AiBase Highlights:】
🌱 The average energy consumption of ChatGPT is 0.3 watt-hours, significantly lower than the previously estimated 3 watt-hours.
🔌 The increase in AI energy consumption is mainly related to future technological advancements and handling more complex tasks.
🏭 OpenAI plans to invest heavily in expanding AI data centers in the coming years to meet the growing power demands.
7. Dongfeng Nissan N7 first vehicle to integrate DeepSeek large model, achieving intelligent human-machine interaction
The new Dongfeng Nissan model N7 has successfully integrated with the DeepSeek-R1 deep reasoning large model, becoming the first joint venture car company to adopt this technology. The N7's innovations in intelligence and human-machine interaction enhance user experience by providing personalized services. Its unique design and powerful intelligent driving system showcase Dongfeng Nissan's commitment to technological innovation, providing a new direction for the future of automotive development.
【AiBase Highlights:】
🌟 The N7 is the first joint venture car to integrate the DeepSeek-R1 large model, enhancing the naturalness and smoothness of human-machine interaction.
🚀 Equipped with the Qualcomm Snapdragon 8295P processor, supporting high-speed navigation and full-scene intelligent parking functions.
💡 Unique design featuring frameless doors and customizable lighting systems, providing personalized visual effects.
8. Thomson Reuters wins landmark AI copyright case in the US
Recently, a US court ruled in favor of Thomson Reuters in a copyright case involving AI technology. This ruling not only provides legal support for Thomson Reuters' copyright protection but also sets an important precedent for other content creators. The court recognized that Thomson Reuters' data is protected by copyright, and unauthorized use constitutes infringement, prompting generative AI companies to be more cautious in their data usage to ensure compliance with copyright laws.
【AiBase Highlights:】
📌 Thomson Reuters won a copyright case involving AI technology in a US court, which has significant legal implications.
⚖️ The court ruled that its data is protected by copyright, and unauthorized use constitutes infringement.
🔍 This case provides an important legal precedent for generative AI companies and content creators in the future.
9. YouTube launches new AI features: automatic dubbing, age recognition, and more
YouTube CEO Neal Mohan emphasized in his annual letter that artificial intelligence is one of the company's key strategies for 2025. YouTube is rolling out multiple AI tools aimed at helping creators enhance content creation efficiency, including automatic dubbing and user age recognition features. These tools will enable creators to easily translate videos and provide personalized recommendations, although the introduction of AI has also raised some concerns.
【AiBase Highlights:】
🌟 YouTube CEO views AI as a key strategy for the company in 2025, launching several AI tools for creators.
🎤 The new automatic dubbing feature will help creators easily translate videos, expected to be fully launched this month.
👶 YouTube will use machine learning technology to estimate user age for more personalized content recommendations.
10. Hisense TV integrates DeepSeek, enhancing AI voice dialogue functionality
Hisense Group announced that its self-developed Xinghai large model has successfully integrated with DeepSeek, aiming to enhance user interaction experiences in smart living and smart city domains. Through this integration, Hisense's intelligent technology capabilities have been significantly enhanced, allowing users to easily activate dialogue functions through remote control or voice commands, simplifying operational steps. Additionally, Hisense has launched a series of smart home assistants to improve the effectiveness of intelligent interactions.
【AiBase Highlights:】
🌟 Hisense TV integrates DeepSeek, allowing users to activate AI dialogue functions via remote control or voice.
🔧 The upgrade of Hisense's Xinghai large model capabilities allows AI to better understand user needs.
🏙️ In smart city applications, Hisense has built an intelligent traffic knowledge base to improve information retrieval efficiency.
