#AI

SoftBank and Intel are partnering to develop an AI-specific memory chip, Saimemory, aiming to reduce AI computing power consumption by up to 50%. This stacked DRAM chip, utilizing an innovative wiring architecture, promises improved energy efficiency and supports sustainable AI data centers. With an estimated $64 million investment, a prototype is expected within two years, targeting commercial deployment in the late 2020s, starting with SoftBank’s AI data centers. Other research institutions and the government are expected to offer support as the project progresses.

Meta plans to fully automate AI-powered ad creation for brands by the end of next year, leveraging AI to generate visuals, videos, and text based on product images and budget goals. This initiative aims to further strengthen Meta’s advertising business, which accounts for over 97% of its revenue. Simultaneously, Meta is shifting towards AI for risk assessment, targeting AI to handle up to 90% of related tasks.

Following remarks at the Shenzhen auto show comparing product performance, a tech executive ignited online debate on user experience and sales discrepancy. Cryptic responses and fan base clashes ensued, prompting calls for industry leaders to foster mutual respect and collaboration. A commentator emphasized constructive dialogue over public disparagement, recognizing the contributions of Chinese tech companies to innovation.

Stanford researchers have unexpectedly created AI-generated CUDA kernels that surpass human-optimized versions. The AI-powered kernels demonstrated significant performance gains across various deep learning operations, outperforming PyTorch in several benchmarks, including matrix multiplication and 2D convolutions. This breakthrough, achieved through a novel method involving language reasoning, could redefine kernel optimization and potentially impact deep learning development.

AWS is aggressively expanding its global infrastructure and securing NVIDIA’s AI chips to meet surging demand, anticipating billions in annual AI revenue. While AWS revenue increased in Q1, it lagged behind competitors like Microsoft Azure and Google Cloud. CEO Matt Garman highlighted datacenter expansions, collaboration with NVIDIA on GB200 chips, and willingness to host OpenAI models, advocating for cross-platform partnerships.

FiEE, Inc. (MINM) will resume trading on the Nasdaq on June 2, 2025, after receiving approval. The company, previously known as Minim, is refocusing on leveraging IoT, connectivity, and AI for brand management solutions, including AI-driven content creation for KOLs and brands. CEO Rafael Li highlights the re-listing as validation and emphasizes strategic initiatives for long-term growth.

Honor CEO Li Jian announced the company’s entry into robotics, highlighting a prototype robot capable of 4 meters per second, potentially surpassing competitors like Unitree Robotics. This launch, backed by a $10 billion investment over five years, follows Honor’s “Alpha Strategy” focusing on AI and robotics. Subsidiaries like Guiyang and Harbin Xingyao Terminal Co., Ltd. have been established to support their robotics endeavors, indicating a growing trend in the smart hardware market.

DeepSeek unexpectedly updated its R1 model, rather than releasing R2. This trial upgrade, accessible via their website and apps, maintains the existing API. Early tests of the open-source R1 (0528) version show significant improvements, especially in programming, now rivaling international standards. The updated model excels in varied tasks from web design to text processing, sometimes outperforming other models like Claude 4.

CNPC launched the Kunlun large language model, a 300-billion-parameter AI, marking a significant advancement for the energy giant. Developed with partners including China Mobile and Huawei, it’s the first of its kind approved for China’s energy sector. The model supports 100 application scenarios and includes a visual AI model built with Huawei, leveraging Ascend hardware. The Kunlun AI Middle Platform, operating on a China-made infrastructure, supports training and deployment, with computing power reaching 1950P.

G&M Direct Hire, a UK car hire firm, achieved significant operational improvements using Samsara’s Connected Operations® platform. They realized £1.8M in accident-related savings, recovered £3M from stolen vehicles, and saved £400K by avoiding excess mileage. The platform provides real-time data, boosting safety, efficiency, and aiding a transition to an electric vehicle fleet.