Recently, Meitu Image Research Institute (MT Lab) and Beijing Jiaotong University jointly proposed a technique called MEMatte (Memory Efficient Matting), which is a super high-resolution matting technology successfully selected for the top-tier conference in the field of artificial intelligence, AAAI 2025. The main highlight of the MEMatte technology is that it is a memory-friendly natural image matting framework that effectively reduces the computational overhead of the model. This innovation allows for operations in memory-constrained environments, such as commercial graphics cards and edge devices.

Traditional video understanding models face numerous challenges when processing long videos, including the complexities of understanding the extended context. Although considerable research has been conducted to enhance video understanding capabilities, effectively overcoming the issues of low training and inference efficiency remains difficult. To address these challenges, the research team utilized HiCo technology to compress redundant parts of video information, significantly reducing computational demands while retaining key information. Specifically, HiCo achieves hierarchical compression of the video by segmenting long videos into shorter clips, thereby reducing processing time.

In multimodal tasks, visual language models (VLMs) play a crucial role, such as in image retrieval, image captioning, and medical diagnosis. These models aim to align visual data with language data for more efficient information processing. However, current VLMs still face significant challenges in understanding negation. Negation is critical in many applications, such as distinguishing between "a room without windows" and "a room with windows." Despite significant progress made by VLMs, existing models fall short when it comes to handling negated statements.

Recently, researchers at Concordia University launched a new research project called 'Abundant Intelligences,' aimed at reexamining the development direction of artificial intelligence (AI). The project points out that the current development model of AI has inherent biases against non-Western conceptions of intelligence, particularly in relation to indigenous cultures. The goal of the project is to integrate indigenous knowledge systems to construct an inclusive and comprehensive concept of intelligence and intelligent behavior, thereby promoting the development of future technologies.

At the recent Qianfan Think Tank Life Sciences Forum, Baidu Smart Cloud launched the new biocomputing model tool HelixFold 3. This tool was co-developed by Baidu's PaddleHelix team and the CHPC High-Performance Computing Platform team, aiming to provide more accurate, efficient, and convenient solutions for life sciences research. The release of HelixFold 3 marks another technological breakthrough in the field of biocomputing. This model is particularly focused on RNA structure prediction and covalent modification.

Recently, a research team from New York University published a study revealing the vulnerabilities of large language models (LLM) in data training. They found that even a tiny amount of false information, constituting just 0.001% of the training data, can lead to significant errors in the entire model. This finding is particularly concerning in the medical field, where inaccurate information can directly impact patient safety. Image credit: Image generated by AI, image licensed by Midjourney. The researchers published their paper in the journal Nature Medicine.

On January 11, 2025, Beijing JD Century Trade Co., Ltd. held the 2025 JD Health Annual Doctor Celebration and Smart Medical Conference in Beijing. At the conference, JD Health announced the launch of its full range of large model products "AI Jingyi," based on its online comprehensive scenarios, as well as the industry's first large model product "JD Zhuoyi" aimed at full-scene applications in hospitals. The goal is to leverage the technological capabilities of the "Jingyi Qianxun" medical large model to create a service system for large model products covering both online and offline medical health scenarios, enhancing the patient experience and supporting doctors in clinical practice.

The World Economic Forum recently released the "Future of Jobs Report 2025," revealing an intriguing contradiction: while artificial intelligence (AI) is seen as a technology that will replace certain jobs, global companies are more inclined to retrain existing employees rather than resort to layoffs. The report surveyed 1,000 companies worldwide, and the results showed that about 41% of companies plan to cut positions that AI can handle, while 77% focus more on training current employees in AI skills. This trend highlights that many companies wish to adapt to AI through retraining.

According to a report by Late Auto, BYD is accelerating its research and development in the field of artificial intelligence. Nio Pengcheng, the former chief technology expert of Zero One Technology, joined BYD's Advanced Technology Research Center in November 2024. He will primarily be responsible for work related to AI supercomputing and may be involved in the development of AI applications. With nearly 12 years of rich experience in AI computing, Nio Pengcheng brings new vitality to BYD's technological advancements. To speed up the pace of technology research and development, BYD has recruited experienced professionals from large model companies.

On the occasion of Baidu's 25th anniversary, company founder Robin Li sent a letter to all employees, reflecting on Baidu's development history and looking ahead to the future. In the letter, Robin Li emphasized that Baidu has always been at the forefront of technological development, firmly believing that technological innovation is the core competitiveness of the company. Robin Li pointed out that throughout its history, Baidu has consistently invested over 20% of its funds in R&D, aiming to transform cutting-edge technology into practical products to benefit more people. He stressed that the large-scale application of technology is key to realizing its value and achieving significant technological breakthroughs and disruptive innovation.

As generative artificial intelligence (Gen AI) technology becomes a hot topic in the smartphone industry, Counterpoint Research recently conducted a survey of consumers in seven countries. The survey included over 25,000 participants from the United States, Canada, the United Kingdom, France, Germany, Poland, and Japan. The results showed that only 32% of respondents were familiar with Gen AI, with the highest awareness in the U.S. at 72%. In contrast, Japan had the lowest awareness at only 7%.

A recent study suggests that artificial intelligence (AI) has the potential to surpass human intelligence in the future. The research delves into the complexities of AI and the human brain and proposes a new theoretical framework, indicating that by applying AI technology at the cellular level of neuroscience, AI could infinitely approach the functionality of the human brain, potentially even surpassing human intelligence levels. The human brain is widely regarded as one of the most complex systems in the universe, while artificial intelligence is seen as one of the most important technologies in human history. A core question looms: Will artificial intelligence ultimately surpass human intelligence?

Hawaii public schools recently piloted an artificial intelligence project, the results of which left a deep impression on education leaders and even moved them. Six teams of 80 students from elementary, middle, and high schools across the islands were tasked with identifying challenges faced by their peers and creating an AI program or "chatbot" to help find solutions. The project was co-created by Ian Kitajima, president of the Pacific International Center for High Technology Research. Hawaii Public Radio (HPR) interviewed Kitajima to learn more.