Odaily Planet Daily News OpenAI has released its latest flagship model, GPT-4o, which can perform real-time inference on audio, visual, and text data. It is designed to be a personalized voice interactive assistant with human-like, supernatural, and ultra-low latency capabilities. According to the official website of OpenAI and the official account of the X platform, the "o" in GPT-4o stands for Omni, representing a step towards more natural human-machine interaction. It accepts arbitrary combinations of text, audio, and images as input, and supports generating arbitrary combinations of text, audio, and image outputs. It can respond to audio input within 232 milliseconds on average, which is similar to human reaction time in conversations. In terms of English and code, its performance is comparable to GPT-4 Turbo, with significant improvements in non-English language texts. Additionally, the API speed is faster and costs 50% less. Compared to existing models, GPT-4o excels in visual and audio understanding. Text and image input will be launched today in API and ChatGPT, while voice and video input will be launched in the coming weeks.

According to the report, this month, with the decrease in the heat of the Spring Festival, the number and intensity of activities of some top products on mobile and client platforms have decreased, and the performance of new products has been unable to support the increase. As a result, the domestic mobile client game market has shown a month-on-month decline. In March 2024, the revenue of the Chinese game market was 23.417 billion yuan, a month-on-month decrease of 5.86% and a year-on-year increase of 7.18%. Among them, the actual sales revenue of the Chinese mobile game market was 16.953 billion yuan, a month-on-month decrease of 7.13% and a year-on-year increase of 8.75%. The actual sales revenue of the Chinese client game market was 5.528 billion yuan, a month-on-month decrease of 2.02% and a year-on-year decrease of 0.39%. The actual sales revenue of Chinese self-developed games in the overseas market was 1.427 billion yuan, a month-on-month increase of 5.98% and a year-on-year increase of 11.34%.

1. The United States and China announced that they will hold the first meeting of the U.S.-China Intergovernmental Dialogue on Artificial Intelligence. 2. The world's first full-scale humanoid robot "Tiangong" with pure electric drive for anthropomorphic running was released3. Apple has restarted negotiations with OpenAI to add new AI features to new products. 4. Musk: TSL will invest about $10 billion this year in AI training and inference in the automotive field. 5. The U.S. government reportedly set up an AI Security Committee, whose members include executives from tech giants such as Jensen Huang and Sam Altman. 6. China's version of Sora-level video model Vidu released: it can generate up to 16 seconds and up to 1080p video. 7. Nvidia Jensen Huang says AI won't completely replace human jobs. 8. The EMO model will be fully launched on the Tongyi App for free, and the cooperation with enterprise customers will be opened as soon as possible. 9. Tsinghua University established the School of Artificial Intelligence, and the first dean was Academician Yao Chizhi, a Turing Award winner. 10. Danghong Technology released the longest audio-visual model BlackEye. 11. General k question Open Source k billion-level parameter model.

Golden Ten Data on April 28, according to Danghong Technology, at the "2024 Zhongguancun Forum - UHD Audiovisual Technology Innovation and Development Forum" held on the afternoon of April 27, Danghong Technology and Beijing Economic and Technological Development Zone jointly released the BlackEye large model base and application scenarios. BlackEye integrates long kinds of Depth neural network components, including Transformer, Diffusion and other components, through text, image, video and audio, 3D model and other long modal encoding, decoding, long modal latent short alignment, long modal language reasoning and generation and other technologies, to achieve inference and prediction generation between different modal information.