Machine learning, especially deep learning, represents a new paradigm of programming and problem-solving.
Traditional programming is logic-driven, encoding knowledge explicitly as rules and algorithms; machine learning is data-driven, with models learning implicit knowledge and patterns from data.
If we take technology solving problems as the primary route:
The core goal of programming in the internet era: building connections, transmitting information, optimizing processes, and providing services. Programming revolves around clear logic and goals, with relatively clear objectives and easily measurable outcomes.
Under this route, the core driving force of AI is to solve more complex, fuzzier problems that traditional programming struggles to address. Beyond this route, AI’s initial development also had other objectives such as creating thinking machines and exploring the essence of learning. Today, the AGI narrative is also becoming a clear route.
However, the AGI narrative at a macro level can overshadow the actual results of problem-solving. We may overlook, undervalue, or overestimate this new paradigm—the new AI-centric paradigm that changes how knowledge is acquired, expressed, and applied, giving rise to “Nexus” (From Company to Nexus: A New Paradigm for Human-AI Collaboration), which represents the possibility of deep integration between AI and humans.
I firmly believe:
We urgently need a more sober and balanced perspective. We must maintain enthusiasm for exploration and forward-thinking regarding long-term goals like AGI, while also fully recognizing and highly valuing current AI technologies represented by deep learning as a brand new, data-driven, problem-solving paradigm with its own independent value and revolutionary significance.
This new paradigm is enhancing our ability to understand and transform the world in unprecedented ways. It is not merely some intermediate stage towards AGI (though its research outcomes may contribute to AGI)—it itself is defining a new era—an “AI-first” era of intelligent augmentation. The future collaborative form of “Nexus” that I define is precisely the inevitable product of the continuous deepening of this paradigm.
机器学习，尤其是深度学习，代表了一种新的编程和问题解决范式。
传统编程是逻辑驱动的，编程将知识显式地编码为规则和算法；机器学习是数据驱动的，模型从数据中学习到隐性的知识和模式。
如果以技术解决问题为主要路线：
互联网时代编程的核心目标：构建连接、传递信息、优化流程、提供服务。编程围绕着明确的逻辑与目标进行，目标相对清晰、成果易于衡量。
这种路线下 AI 的核心驱动力是解决那些传统编程难以解决的、更复杂、更模糊的问题。除这条路线外，AI 最初的发展也有打造会思考的机器、探讨学习本质等其他目标，当下，AGI 叙事也是一种路线且越来越清晰。
恰恰 AGI 叙事在宏观上会覆盖掉真正解决问题的结果，我们也许会忽略、看低、夸大这种新范式，即以 AI 为主的新范式，改变了知识的获取、表达和应用方式，催生了”Nexus”(From Company to Nexus: A New Paradigm for Human-AI Collaboration)，即 AI 与人深度融合的可能。
我深以为然：
我们亟需一种更为清醒和平衡的视角。既要对 AGI 这样的长远目标保持探索的热情和前瞻的思考，更要充分认识和高度评价当前以深度学习为代表的 AI 技术，作为一种全新的、以数据为驱动、以解决问题为导向的范式，其本身所具有的独立价值和革命性意义。
这种新范式正在以前所未有的方式增强我们认识世界、改造世界的能力。它不仅仅是通往 AGI 的某个中间阶段（尽管其研究成果可能对 AGI 有贡献），它本身就在定义一个新的时代——一个”AI 为主”的智能增强时代。我定义的”Nexus”未来协作形态，正是基于这一范式不断深化的必然产物。