Search AI Products and News

A collaborative team from Ant Group, Zhejiang University, University of Liverpool, and East China Normal University presented ScaleOT, a novel cross-domain fine-tuning framework, at AAAI 2025. This framework achieves a 50% improvement in privacy protection and a remarkable 90% reduction in computational cost compared to knowledge distillation, without compromising model performance. ScaleOT offers a highly efficient and lightweight solution for cross-domain fine-tuning of billion-parameter models. The innovative nature of the paper secured it an oral presentation at AAAI.

The AI field is abuzz with the recent news of Moonshot's open-sourcing of its new optimizer, Muon, which boasts a computational efficiency twice that of traditional AdamW. This release coincides with DeepSeek's upcoming release of several codebases, generating significant industry attention and discussion. The Muon optimizer was initially proposed by OpenAI researchers Keller Jordan et al. in 2024 and showed excellent performance in training small-scale models. However,

As businesses increasingly adopt large language models (LLMs), improving the accuracy of model knowledge and reducing hallucination phenomena has become a significant challenge. Researchers at Meta AI have introduced a 'Scalable Memory Layer' in a recent paper, which may provide a solution to this issue. The core idea of the scalable memory layer is to add more parameters to the LLMs without increasing computational resources during inference, thereby enhancing their learning capabilities. This architecture is suitable for storing large amounts of factual knowledge while maintaining...

For a long time, efficiently generating high-quality, wide-angle 3D scenes from a single image has been a challenge faced by researchers. Traditional methods often rely on multi-view data or require time-consuming per-scene optimization, and they struggle with background quality and reconstruction of unseen areas. Existing technologies, when dealing with single-view 3D scene generation, often produce errors or distortions in occluded areas due to insufficient information, along with blurry backgrounds and difficulties in inferring the geometric structure of unseen regions. While regression-based models can synthesize new views in a feedforward manner, they struggle with complex...

On December 24, the Shanghai Municipal Government Office announced the 'Work Plan for the Development of Medical Artificial Intelligence in Shanghai (2025-2027)'. The plan clarifies the main tasks and goals for the next three years, aiming to respond to the national strategic deployment of 'AI + Action'. By promoting deep integration of medical innovation and artificial intelligence, it seeks to transform Shanghai into a globally influential hub for innovation, application demonstration, and industrial development in medical AI technology.

Recently, Rony Abovitz, the founder of Magic Leap, has once again captured public attention with his new company SynthBee, which has just completed a $20 million funding round. Located in Fort Lauderdale, Florida, the same place where Magic Leap rose to fame and sparked excitement around mixed reality computing, this funding aims primarily to expand the team and advance its unique computational intelligence platform. Image credit: Photo by AI generation.

According to Xinhua News Agency, the Madrid Zoo Aquarium recently announced that a unique AI and synthetic biology-developed zookeeper, Irenia, has joined the giant panda support team, providing relevant information for the pandas 'Jinxi' and 'Zhuyu,' who arrived from China in April this year. Irenia is a digital persona that previously appeared on the zoo's TikTok channel, offering educational insights about giant pandas. It is now featured on a large screen behind the panda exhibit, providing real-time information to visitors.

Recently, a research team from Shanghai Jiao Tong University and Harvard University introduced a novel model fine-tuning method — LoRA-Dash. This new approach claims to be more efficient than existing LoRA methods, particularly in the fine-tuning of specific tasks, achieving the same results with an 8 to 16 times reduction in the number of parameters. This is undoubtedly a major breakthrough for fine-tuning tasks that require substantial computational resources. In the context of the rapid development of large-scale language models, the demand for fine-tuning specific tasks is steadily increasing. However, fine-tuning often

Israeli artificial intelligence startup CytoReason recently announced the completion of an $80 million funding round, with investors including industry giants such as Nvidia, Pfizer, OurCrowd, and Thermo Fisher Scientific. The company plans to use the funds to accelerate the development of its AI platform for disease modeling and drug discovery.Founded in 2016 and headquartered in Tel Aviv, CytoReason's AI platform extracts insights from human disease data to create computational disease models

Recently, Apple's AI team, in collaboration with institutions such as the University of Washington, has introduced an open-source language model named DCLM. This model boasts 700 million parameters and was trained using a staggering 2.5 trillion data tokens, aiding us in better understanding and generating language. So, what exactly is a language model? In simple terms, it is a program capable of analyzing and generating language, assisting us in various tasks such as translation, text generatio

Evo is a genomic foundation model capable of multi-modal and multi-scale learning. Evo can generate genomic-scale DNA sequences. Evo goes beyond single molecules and short sequences, allowing for the design of synthetic multi-component biological systems. Evo can predict essential genes in bacteria or phages at the whole genome scale.

The Colombian Ministry of ICT will invest 10 billion Colombian pesos (approximately $25.2 million) to establish 300 artificial intelligence innovation centers, focusing on computation, gaming, and data technology. The first innovation center will be located in the municipality of Quinchía in the Risaralda department, aimed at enhancing local connectivity and advancing AI projects. This initiative is part of a computational education program promoted by ICT Minister Mauricio Lizcano, designed to strengthen artificial intelligence and computational capabilities across Colombia. The total investment is $250 million.

Meta has released the Llama2-Long model, which does not increase computational demand when processing long texts while still maintaining excellent performance. The model employs continuous pre-training, improved positional encoding, and data mixing strategies to reduce computational overhead by up to 40%. It performs outstandingly on both long and short tasks, surpassing other long-context models and has the potential to revolutionize the field of natural language processing. The model shows significant improvements in encoding, mathematical, and knowledge-intensive tasks, even surpassing GPT-3.5. The release of the Llama2-Long model provides a powerful solution for handling long texts.