Search AI Products and News

OpenAI recently released its latest reasoning models, o3 and o4-mini, marking a significant breakthrough in the field of artificial intelligence. These models not only surpass previous versions in reasoning capabilities but also achieve image reasoning for the first time, integrating visual information directly into the thinking process. o3, hailed as a "genius-level" model, particularly excels in tasks such as programming and mathematics, achieving an accuracy rate of 87.5%. The newly released o3 and o4-mini models demonstrate exceptional performance in multimodal processing, possessing...

Skywork AI's research team recently unveiled SkyReels-A2, a groundbreaking video generation framework pushing the boundaries of controllable video synthesis. This Element-to-Video (E2V) framework generates natural videos from text prompts by composing diverse visual elements (characters, objects, backgrounds), maintaining high fidelity to reference images. SkyReels-A2's core lies in its sophisticated data processing pipeline. The research team designed a comprehensive data construction pipeline.

Artificial intelligence is rapidly transforming protein discovery and design in biotechnology. Recently, a collaborative team from UC Berkeley and Caltech developed ProteinDT, a novel multimodal framework leveraging text descriptions to aid in protein design. This innovative approach integrates protein sequence and structural information with a wealth of biological knowledge represented in text, opening a new chapter in protein design.

Accurate power generation forecasting and supply-demand management are crucial for the rapid development of the new energy industry. On March 26, Ant Group released its new energy power time series large model – EnergyTS. This model aims to help companies optimize their business strategies and improve operational efficiency through high-precision data analysis. EnergyTS demonstrates particularly strong performance in photovoltaic scenarios, surpassing Google's TimesFM-V2.0 and Amazon's Chronos-Lar in power generation prediction accuracy.

Beijing Shengshu Technology Co., Ltd. announced that its AI video generation platform, Vidu, has secured the rights to adapt seven hit online novel IPs into concept shorts. Vidu will leverage AI technology to empower China's animation and short-form drama market, driving the adaptation of online literature into film and television into a new phase of development.

Microsoft, the US tech giant, and Inait, a Lausanne-based startup, announced a strategic partnership on Tuesday to co-develop an innovative AI model inspired by the mammalian brain. According to a joint announcement, Microsoft plans to leverage Inait's unique technology to further expand its AI product portfolio for enterprise clients. Inait's technology is the culmination of 20 years of brain research. CEO Richard Frey highlighted the core principle: the only proven form of intelligence exists.

Recently, the Daye City People's Court in Hubei Province issued a verdict in a landmark case involving the use of artificial intelligence (AI) technology to write and profit from pornographic novels. Defendant Ke Mou was sentenced to ten months imprisonment and a fine of 5,000 yuan for the crime of producing, selling, and disseminating obscene materials for profit, and ordered to return his illegal gains. According to the prosecution, from November 2022 to March 2023, Ke Mou, a college-educated online literature author, used an AI program to write pornographic novels and published them on overseas pornographic websites, simultaneously...

Large Language Models (LLMs) are constantly evolving in the field of artificial intelligence. Researchers from Carnegie Mellon University (CMU) and HuggingFace recently introduced a new method called Meta Reinforcement Fine-Tuning (MRT). This method aims to optimize the computational efficiency of LLMs during testing, particularly excelling in solving complex reasoning problems. Studies show that existing LLMs struggle with...

Analyzing medical image data has always been a complex and laborious process. Recently, researchers at Weill Cornell Medicine developed a novel artificial intelligence system called LILAC (Learning-based Inference of Longitudinal Image Changes) that can efficiently and accurately analyze and detect changes in medical images over time. This research, published on February 20th in the Proceedings of the National Academy of Sciences, showcases LILAC's broad application potential in various medical scenarios. Traditional medical image analysis methods often require extensive customization and pre-processing.

Researchers from the Tübingen University's ELLIS Unit, the University of Maryland, and Lawrence Livermore National Laboratory have developed Huginn, a novel language model with a recursive architecture that significantly enhances reasoning capabilities. Unlike traditional models, Huginn doesn't require specialized 'reasoning chain' training; it can autonomously reason within the neural network's 'latent space' before outputting results. The research team developed Huginn, a novel language model with a recursive architecture that significantly enhances reasoning abilities. Unlike...

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...

Recently, Oute AI released a novel text-to-speech synthesis method called OuteTTS-0.1-350M. This method utilizes pure language modeling without the need for external adapters or complex architectures, offering a simplified TTS approach. OuteTTS-0.1-350M is based on the LLaMa architecture, using WavTokenizer to directly generate audio tokens, making the process more efficient. The model features zero-shot voice cloning capability, requiring only a few seconds of reference audio.