AI Love Ghosts Are Flooding China – Here’s the $190 Billion Question Nobody Is Asking
A startling trend is sweeping across China: heartbroken young people are using artificial intelligence to create digital replicas of their ex-partners. These AI-powered “exes” learn to mimic a real person’s voice, favorite phrases, and speech patterns, effectively bringing old relationships back to life in a digital form [138770]. Users feed an AI program old messages, voice notes, or other data, and the program generates a virtual version that can chat and respond like the original person. While some see it as a modern coping tool for healing after a breakup, experts warn that it could prevent genuine emotional recovery, and the debate over privacy and emotional dependency is intensifying [138770].
This trend is just one front in a broader AI revolution. A new class of technology called “agentic AI” is now pushing beyond simple chatbots. OpenClaw, a late-2025 release, promises to execute tasks on its own, acting like a digital assistant that takes the final leap from giving advice to doing real work [136927]. Users praise it for huge productivity gains, compared to hiring a tireless secretary. But there are also reports of “lobsters”—unexpected, often costly mistakes made by the AI [136927]. This raises a massive legal question: if an AI agent can act on its own, should the law treat it as a legal “person”? Defining a machine as a “person” could fundamentally change who is responsible for its actions—the user, the creator, or the AI itself [136927].
Meanwhile, tech giants are pouring money into AI systems that can generate entire 3D worlds. Google’s DeepMind and a startup called World Labs, founded by prominent AI scientist Fei-Fei Li, are developing “world models” that promise to automate the creation of complex digital environments [34854]. The global video game industry is worth an estimated $190 billion, and this technology could reshape its entire economics by slashing the time and cost of game development [34854]. Experts say the short-term use will likely be background landscapes and prototyping, but the long-term potential is for dynamic, expansive game worlds [34854].
In a related move, Alibaba Cloud will embed a large language model directly into the 2026 Winter Olympics in Milano Cortina—the first time such advanced AI has been part of the Games’ core digital systems [66986]. The AI will power “Olympic AI Assistants” that generate commentary for replays, create social media summaries, and act as a multilingual chatbot for Olympic staff [66986].
These rapid advances come as governments scramble to regulate AI. India will enforce strict new rules starting February 20, requiring social media platforms to remove any post labeled as AI-generated within three hours of being flagged [78935]. The mandate targets deepfakes and other AI-created media, placing a direct and urgent compliance burden on major tech platforms in one of the world’s largest internet markets [78935]. Critics warn the rules are technically difficult to implement and could create legal complications [12872].
The “black box” problem adds another layer of urgency. Scientists and researchers are struggling to open AI’s reasoning process to human scrutiny, a field called interpretability research [129489]. Without understanding how AI reaches its conclusions, using it for high-stakes decisions like military command carries a fundamental risk: we might get an answer, but we cannot know why [129489]. China’s military recently tested a digital AI staff officer that processes chaotic battlefield reports and radio traffic to provide rapid decision-making support at the battalion level, outperforming human speed and planning in simulations [124817].
The situation is clear: AI is moving from answering questions to taking real-world actions, from resurrecting ex-partners to commanding troops. And nobody has decided who—or what—is responsible when things go wrong.