Looking Ahead to 2026
打开互动全文版（逐段中英对照 + 图/公式 + 论文问答）→关于技术进步及其现实影响的笔记
Notes on advances in technology and real-world impact
关于技术进步和实际影响的说明
Notes on advances in technology and real-world impact
回顾过去一年并展望未来，毫无疑问 2026 年将是 AI 的关键一年。是的，又是关键一年。但这次在几个显著方面有所不同。
As I reflect on the past year and look toward the one ahead, there’s no question 2026 will be a pivotal year for AI. Yes, another one. But this moment feels different in a few notable ways.
我们已经度过了最初的发现阶段，正在进入广泛扩散的阶段。我们开始区分“奇观”与“实质”。现在我们对技术走向有了更清晰的认识，但更困难也更重要的问题是如何塑造其对世界的影响。
We have moved past the initial phase of discovery and are entering a phase of widespread diffusion. We are beginning to distinguish between “spectacle” and “substance”. We now have a clearer sense of where the tech is headed, but also the harder and more important question of how to shape its impact on the world.
我们仍处于马拉松的开端。许多事情仍不可预测。在这种“模型过剩”的情况下，能力超越了当前将其用于产生实际影响的能力，以下是我们仍需做对的事情：
We are still in the opening miles of a marathon. Much remains unpredictable. Amidst this “model overhang,” where capability is outpacing our current ability to use it to have real world impact, this is some of what we still need to get right:
1. 一个新的概念，它演进自“思维的自行车”，使我们始终将 AI 视为人类潜力的支架而非替代品。重要的不是任何给定模型的能力，而是人们如何选择应用它来实现自己的目标。我们需要超越“垃圾”与“精致”的争论，发展一种新的“心智理论”平衡，考虑到人类在相互关联时配备了这些新的认知放大器工具。这是我们需要辩论和回答的产品设计问题。
1. A new concept that evolves “bicycles for the mind” such that we always think of AI as a scaffolding for human potential vs a substitute. What matters is not the power of any given model, but how people choose to apply it to achieve their goals. We need to get beyond the arguments of slop vs sophistication and develop a new equilibrium in terms of our “theory of the mind” that accounts for humans being equipped with these new cognitive amplifier tools as we relate to each other. This is the product design question we need to debate and answer.
2. 在将 AI 部署到现实世界影响时，我们将从模型演进到系统。我们在如何持续驾驭模型能力的指数增长，同时应对其“锯齿状”边缘方面学到了很多。我们现在进入了一个阶段，构建丰富的支架来编排多个模型和智能体；考虑记忆和权限；实现丰富且安全的“工具使用”。这是我们必须继续构建的工程复杂性，以便在现实世界中从 AI 中获得价值。
2. We will evolve from models to systems when it comes to deploying AI for real world impact. We have learned a lot in terms of how to both keep riding the exponentials of model capabilities, while also accounting for their “jagged” edges. We are now entering a phase where we build rich scaffolds that orchestrate multiple models and agents; account for memory and entitlements; enable rich and safe “tools use”. This is the engineering sophistication we must continue to build to get value out of AI in the real world.
3. 最后，我们需要做出深思熟虑的选择，如何将这项技术作为解决人类和地球挑战的方案扩散到世界。为了让 AI 获得社会许可，它必须具有现实世界的评估影响。我们在稀缺能源、算力和人才资源上的应用选择将至关重要。这是我们需要建立共识的社会技术问题。
3. And lastly, we need to make deliberate choices on how we diffuse this technology in the world as a solution to the challenges of people and planet. For AI to have societal permission it must have real world eval impact. The choices we make about where we apply our scarce energy, compute, and talent resources will matter. This is the socio-technical issue we need to build consensus around.
最终，进步最有意义的衡量标准是我们每个人的成果。这将是一个混乱的发现过程，就像所有技术和产品开发一样。计算史上的核心一直是赋能个人和组织实现更多，AI 必须遵循同样的道路。如果我们做到这一点，它可能成为计算史上最深刻的浪潮之一。这是我希望我们在 2026 年及以后共同推动的。
Ultimately, the most meaningful measure of progress is the outcomes for each of us. It will be a messy process of discovery, like all technology and product development always is. Computing throughout its history has been about empowering people and organizations to achieve more, and AI must follow the same path. If we do that, it can become one of the most profound waves of computing yet. This is what I hope we will collectively push for in ‘26 and beyond.
关于技术进步和现实世界影响的笔记
Notes on advances in technology and real-world impact