During a continuous 12-day live event, OpenAI released the O1 model API to developers on the 9th day and announced significant upgrades to the real-time API, supporting WebRTC technology. Starting from the release date, OpenAI will provide access to the O1 API for developers at usage level 5. Compared to the previous preview version, this updated O1 model API achieves a 60% reduction in thinking costs and adds advanced visual processing capabilities. Meanwhile, GPT-4o has improved in audio processing.

The competition in TV software is shifting towards advertising. TV manufacturers are increasingly keen on leveraging TV operating systems (OS) for ad placements and user tracking, leading to bold new strategies such as advertising technology companies launching TV OS and displaying ads on TV screensavers. TCL, as a TV manufacturer, is set to premiere a series of shorts on its free streaming service. This marks the company's new attempt at monetizing TV users and film production, relying on generative AI and targeted advertising.

In product teams, the backlog of screen recordings and screenshots often goes unnoticed, with video creation being both time-consuming and costly. However, two former Microsoft employees saw an opportunity in this market and decided to launch an AI tool called Lica, which easily transforms screenshots and recordings into engaging tutorials and product videos. This startup was founded in 2023 by Priyaa Kalyanaraman and Purvanshi Mehta, both ex-Microsoft employees.

Recently, the AI research company Epoch AI released an interactive simulator designed to simulate the computational power required to train large language models. Through this simulator, researchers found that while a 2012 outdated graphics card (like the GTX580) could be used to train GPT-4, the cost would be ten times that of modern hardware today. Epoch AI's research indicates that the number of floating-point operations (FLOP) required to train GPT-4 is between 1e25 and 1e26.

Zhou Hongyi, the founder of 360, recently announced that the company plans to launch a brand new search product this month, which will further expand 360's footprint in the field of artificial intelligence applications. Zhou pointed out that 360 AI Search, as one of the largest native AI applications in China, has proven that search is a direct and simple way for users to access and use AI technology. Based on this successful experience, the 360 team is committed to innovating on the existing foundation to create a more accessible and user-friendly AI tool.

Recently, the globally renowned robot vacuum manufacturer iRobot announced it will lay off approximately 105 employees, which accounts for 16% of its total workforce. This news was disclosed in the company's financial report, where iRobot's CEO Gary Cohen stated during the earnings call that this move, while challenging, fundamentally changes the way the company works with its partners, improving the efficiency of robot research and development as well as production. Cohen emphasized that iRobot has already...

Generating high-quality, temporally continuous videos requires a significant amount of computational resources, especially for longer time spans. The latest Diffusion Transformer model (DiTs) has made significant progress in video generation, but its reliance on larger models and more complex attention mechanisms has resulted in slower inference speeds, exacerbating the challenge. To address this issue, researchers at Meta AI have proposed a training-free method called AdaCache to accelerate video DiTs.

Large Language Models (LLMs) like GPT and Llama have sparked a revolution in the field of artificial intelligence, but efficiently training these massive models while aligning them with human values remains a challenge. Reinforcement Learning with Human Feedback (RLHF) has become an important training method for LLMs in recent years, but traditional RLHF frameworks have limitations in flexibility, efficiency, and scalability. To address these issues, the ByteDance Doubao Large Model Team has open-sourced an RLHF framework called HybridFlow.

Beijing Baichuan Intelligent Technology Co., Ltd. recently announced the launch of its one-stop solution aimed at helping enterprises achieve private deployment of large models at a low cost while improving efficiency. This solution includes high-quality general training data across the entire link, two models: Baichuan4-Turbo and Baichuan4-Air, as well as a comprehensive toolchain for domain enhancement, meeting the needs of enterprises in various scenarios.

Recently, the corporate legal industry has witnessed an exciting transformation. The Association of Corporate Counsel (ACC) and Everlaw have jointly released a new report on the impact of Generative AI (GenAI) on the corporate legal industry. The report, titled 'Gen AI and the Future of Corporate Legal Work: Is Your In-House Team Ready?', reveals the rapid application of Generative AI in legal work and its profound effects on the legal profession. Image credit: Image generated by AI, image licensed from Midjourney.