360 Security Technology (SH:601360) Bei Ke Cai Jing·2025-08-07 13:00Core Insights - The 13th Internet Security Conference in Beijing featured a humanoid robot named "Red Coat" as the host, with Zhou Hongyi, founder of 360 Group, delivering a keynote titled "ALL IN AGENT" [1] - Zhou expressed concerns about AI safety, emphasizing the need for a "power switch" for large models to ensure human control over AI capabilities [2][3] AI Safety and Control - The rapid advancement of AI models has led to increased risks, with Zhou advocating for the analysis and monitoring of large models using other large models [2] - Zhou criticized the notion of aligning AI with human values, stating that current large models have capabilities that surpass individual human understanding [2] Evolution of Intelligent Agents - Zhou outlined a clear progression of intelligent agents from L1 (chat tools) to L4 (multi-agent collaborative systems), highlighting the limitations of each level [4] - The L4 level, referred to as "multi-agent swarm collaboration space," allows for flexible grouping and task execution, achieving a success rate of up to 95.4% for complex tasks [5] Competitive Landscape - Zhou acknowledged the intense competition in the AI industry, suggesting that the only sustainable advantage is execution capability and the ability to adapt quickly [6] - The rapid evolution of intelligent agents means that any product's lead time is short, with new concepts quickly being adopted by competitors [6]