02 Mar 2026
AI twins yourself
Digital human twins refer to preserving a person in digital form, including their voice, image, and even their patterns of thinking. In the past, this concept appeared only in science fiction films — for example, in certain Hollywood blockbusters, a device scans someone’s eyes and transfers memories and consciousness into a replica. What once existed only in fiction now seems to be gradually entering reality through AI, and the trajectory is becoming increasingly clear.
Similar to what Palantir Technologies does with its Ontology framework — structuring and feeding all enterprise data into AI to create a digital twin of a company — once a system is fully modeled, it can simulate future outcomes and explore different scenarios. A digital twin becomes a predictive environment. A digital human twin follows the same logic. If enough data about a person is provided — video, voice, language, written text, behavioral traces — AI can train on it and reconstruct a highly realistic simulation of that individual. External traits like voice and appearance are already replicated with striking accuracy. As for thought patterns, they are also learnable structures. With sufficient interaction data and expanded long-term memory, AI could approximate how someone reasons, responds, and makes decisions. It could approach near-complete behavioral imitation and even predict likely next actions.
Extending this logic further, digital twins are not limited to people. With sufficient data, many systems can be simulated: weather, plant growth, supply chains, inventory management — essentially any complex dynamic process. Some argue that financial markets cannot be predicted; otherwise, everyone would win. That objection reflects current data limitations rather than theoretical impossibility. Stock charts represent only one visible dimension. If data coverage expanded to include macroeconomic conditions, corporate fundamentals, insider activity, capital flows, regulatory shifts, and behavioral psychology, predictive models could approach extremely high accuracy. True black swan events might still exist, but their probability would shrink toward the margins. The challenge lies in complexity and completeness of data, not in principle.
Compared with modeling global markets, simulating a person — a loved one, a pet, a historical figure like Albert Einstein, a CFO, or a personal trainer — is a more bounded problem. The variable space is smaller and the system more constrained. It is therefore reasonable to expect that personal digital twins will emerge earlier than fully predictive macro systems. The productivity gains could be enormous, yet the boundary between human and digital identity would become increasingly blurred. You might spend hours speaking to someone who has no biological existence. Influential public figures might turn out to be entirely synthetic. Eventually, one might even question whether one is already inside a digital simulation — perhaps even an NPC in a larger computational system.
Human beings have always pursued immortality, yet it may arrive not through biology but through data continuity. In speculative futures imagined by Liu Cixin, the world becomes a colossal computational structure — no organic matter, only electricity and signals. Consciousness persists in digital form. And then, eventually, even that system collapses. The structure disappears. Plants grow again. The Earth becomes vibrant and alive — but without humanity.
数字孪生人，就是用数字的形式保存你这个人，包括你的声音，图像，甚至思考。这之前只在科幻电影中出现，比如某好莱坞大片，用设备拍一下眼睛就把思想和记忆传给复制人了。这个只在科幻下存在的场景，现在通过AI似乎慢慢走入现实了，甚至越来越清晰了。
和Plantir公司做的Ontology类似，通过将所有的数据交给AI，AI通过训练和理解，将一个公司变成一个数字孪生，现在会发生什么，将来会有什么变化，数字孪生可以给你一个模拟。那么数字孪生人也是一样，只要提供AI这个人足够的数据，包括影像，声音，语言，文字等等，AI完全可以通过训练，将这个人完全模拟出来。声音外貌这些外在的自然不在话下，现在AI模拟的非常完美了，至于思想也是一样，只要与AI进行了足够的交流，并且AI技术上可以有更长的记忆以后，就会和你现在使用Gemini的GEM或者chatGPT的GPT一样方便，能够无限接近于100%模仿真人的思想，预知下一步的动作。
进一步，你也可以推想，不仅仅是数字孪生人，只要有足够的数据，那么很多场景都能被模拟，天气，植物生长，库存管理等，和拟合有点像。有人会提供反对意见比如股市如果被预测，K线能被预测那不是人人都赢麻了，显然是有局限的。没错，不过这也是数据覆盖率不足，比如K线只是一个纬度，当数据覆盖股市的方方面面，宏观环境，单个公司的财务，大股东的操作等等，自然是可以99.99%模拟出走势的，我留了个0.01%因为还是有黑天鹅的存在的。不过当然这是一个极端复杂的系统，不是短期就能够实现的，相对于这些复杂场景，一个亲人，一个宠物，或者一个爱因斯坦，一个财务总监，一个私人教练是更容易被模拟出来。可想而知，此时将达到更大的生产力解放，但是同时人与数字之间的界限也更为模糊了，有一天，你发现和你聊半天的都不是一个真人，广为人知的大V居然不存在，甚至你发现原来你现在就在数字世界里，你自己就是一个NPC。到那时你就和很多科幻片一样了。
人类希望永生，但是没想到是通过这种数字的方式达到永生，真如刘慈欣所设想的未来世界，整个世界就一个大柱子，里面是精密集成的电子设置。没有有机物，只有是电和信号。人物在数字模拟下达到了“永生”，然后一切归于原始，柱子消失了，植物又生长出来了，地球一片生机勃勃，但是没有了人类。
Til next time,
at 00:00
