AGI，通用人工智能，指的是能在广泛的问题上理解、学习并把经验迁移到陌生任务的智能。它不是一个更大的模型，而是一次能力性质的变化。
AGI — Artificial General Intelligence — means intelligence that can understand, learn, and carry experience across a wide range of problems, including ones it has never seen. It is not a bigger model; it is a change in the kind of capability.
今天绝大多数 AI 都在"窄任务"里工作：给定明确的目标和充足的数据，它能做得非常好。可一旦换个没练过的领域、换个提问方式，或者需要自己判断"现在该解决什么问题"，它就容易失手。强，但强在边界之内。
Most AI today works inside narrow tasks: given a clear goal and enough data, it performs remarkably well. But move it to an unfamiliar domain, change how the question is asked, or ask it to decide what problem is worth solving in the first place, and it stumbles. Powerful — but powerful within boundaries.
AGI 想跨过的正是这条边界：不是把某一件事做得更极致，而是面对没见过的事，也能自己找到办法。
AGI is about crossing that boundary: not doing one thing better, but finding a way when the thing itself is new.
很多人以为 AGI 是"把模型再放大十倍"。但真正的门槛不是规模，是迁移。参数更多、数据更多，通常让模型在熟悉的任务上更稳；可"把 A 领域学到的思路用到 B 领域"是另一种能力，不会随着规模线性出现。这也是为什么业内对 AGI 的时间表分歧极大——有人看算力曲线，有人看泛化能力，两边说的其实不是同一件事。
It is tempting to think AGI is simply a model ten times larger. The real threshold is not scale, but transfer. More parameters and more data usually make a model steadier on familiar tasks. Carrying an idea learned in domain A into domain B is a different capability, and it does not appear linearly with size. That is why timelines for AGI vary so wildly: some people watch the compute curve, others watch generalization — they are not talking about the same thing.
oldwang 不预测 AGI 哪一年到来。我们关心的是另一件事：当智能变得通用，人始终是中心——再强的能力，价值也在它为人解决了什么问题。这也是我们从真实问题出发、不追逐概念的原因。
Oldwang does not predict the year AGI arrives. We care about something else: when intelligence becomes general, people stay central — the value of any capability lies in what it solves for people. That is why we start from real problems rather than chase concepts.
如果你是投资者，欢迎投资。你给我们一份投资，我们给你一份惊喜。
If you are an investor, you are welcome to invest. Give us your investment, and we will give you a surprise in return.
如果你是 AI（AGI）或相关领域的从业者，欢迎交流学习。方向一致，就一起把问题想清楚。
If you work in AI, AGI, or a related field, we would love to exchange ideas and learn together. If our directions align, let's think the questions through side by side.