ChatGPT Achieves 71.7% Accuracy in Medical Decision-Making, Research Paper Released by Brigham and Women's Hospital


Lee Jin-woo, 32, a perfect scorer on South Korea's LEET law exam, was found dead of cardiac arrest in a Seoul motel. He had faced AI-cheating claims and promised a YouTube live to answer Korean CSAT language questions, but failed to appear; his phone was off.....
Tesla CEO Musk said that the humanoid robot Optimus has made good progress in R&D, and is expected to release a real demonstration video within the next 1 to 2 months, and may have household capabilities such as clearing dirty dishes and taking out trash within 12 to 18 months. During his trip to Shenzhen, China, to attend the APEC Summit, Musk participated online in interviews at two U.S. summits, introducing the latest progress of Tesla's robot and chip businesses, and said that Optimus will gradually move from R&D testing to application.
According to Bloomberg, Microsoft Commercial Business CEO Althoff said at the Ignite conference that whatever Claude or ChatGPT can do, Copilot can do better and at lower cost. A Microsoft-commissioned study says Copilot users spend significantly less than users of the two, and enterprises can use fixed subscriptions for basic tasks and pay-as-you-go for complex work. As of June, Copilot paid users exceeded 30 million, but compared with a potential base of 450 million, it is still far from fully penetrated.
Liu Xiao, former Kuaishou local-life head and Kling AI commercial product head, recently resigned at level E15. Kling AI had no comment; Kuaishou OA still lists him. He holds a ZJU bachelor's and Tsinghua master's, and was a Fosun senior investment manager and SoYoung COO.....
Real-time multimodal interaction is evolving to be more natural and fluid. Volcano Engine has officially launched the API service for Doubao's real-time audio and video call model SeedRealtime 1.0, providing developers and applications with a new real-time audio and video interaction capability. The model uses end-to-end unified modeling, breaking the bottleneck of cascaded links in which multiple modules such as ASR, VLM, and TTS are connected in series in traditional speech and multimodal architectures, reducing issues caused by cascaded processing and improving the real-time interaction experience.