离开小冰后，李笛重回大模型牌桌

Core Viewpoint - The article discusses the emergence of Nextie, a new company founded by former Xiaoice CEO Li Di, which aims to explore a different approach to AI that focuses on cognitive models rather than larger models or chatbots [4][5][6]. Group 1: Company Overview - Nextie was established on December 7, with plans for a second round of financing expected to reach several tens of millions of dollars [4]. - The founding team includes key members from Xiaoice, such as co-founder Zeng Min and former Intel architect Wang Wenlan [6]. Group 2: New AI Approach - Li Di emphasizes that Nextie will not continue Xiaoice but will pursue the path that Xiaoice could not complete, focusing on a "cognitive model" rather than a larger chatbot [5][6]. - The cognitive model aims to create a network of agents with different cognitive paths that collaborate to solve complex problems, rather than relying on a single, larger model [7][8]. Group 3: Industry Context - The current AI industry consensus prioritizes larger models and longer contexts, but Li Di argues that this approach does not address the essence of intelligence [8][10]. - The article notes a shift in the industry as companies begin to realize that simply being able to answer questions is not enough; the ability to participate in decision-making and provide structured reasoning is becoming crucial [10][11]. Group 4: Challenges Ahead - Nextie faces several challenges, including acceptance in the capital market, technological maturity, and pressure to establish a viable C-end business model [12]. - The article highlights the tension between the need for innovative approaches and the existing market's focus on traditional metrics like model size and speed [12][14]. Group 5: Vision and Philosophy - Li Di's vision for Nextie is to provide users with a team-like intelligence rather than just enhancing model capabilities, focusing on transparency, independent cognition, and the ability to challenge different perspectives [14][15]. - The approach is characterized by a long-term perspective, acknowledging that the development of such technology may take time and may not be immediately understood or validated by the market [15][16].