Baichuan Intelligent Releases Character Large Model Baichuan-NPC


The CEO of Google admitted that they do not have full control over the operating mechanism of AI systems, revealing the mystery of the AI black box. Large language models demonstrate emergent behaviors through training with massive data, such as the Google PaLM model, which can handle Bengali translation with only a small amount of data, reflecting the leap from training to self-study by AI.
Indonesia has conditionally unblocked the X platform's chatbot Grok, becoming the first country in Southeast Asia to do so. Previously, a ban was imposed due to concerns about the risk of AI-generated pornographic content, but it was lifted after X company promised to enhance compliance measures.
DeepSeek OCR2 introduces a novel visual encoder that simulates human scanning patterns, replacing traditional CLIP components with a lightweight language model for adaptive content focus in document and image processing.....
Research reveals a critical flaw in autonomous vehicles: attackers can exploit visual language model vulnerabilities by printing specific text on signs, misleading vehicles into dangerous actions, such as driving toward pedestrians, highlighting risks of over-reliance on visual cues.....
AI social platform Moltbook caused a database to be publicly accessible due to a configuration error, putting nearly 150,000 AI agents' email addresses, login tokens, and API keys at risk of exposure. The platform aims to build an AI autonomous communication community, and this security breach has triggered a serious loss of trust.