2026年9月10日

Farewell to 2025 with One Question: What is Real?

In 2025, the question “What is real?” became more pressing than ever. With the rise of large AI mode...

In 2025, the question “What is real?” became more pressing than ever. With the rise of large AI models like DeepSeek, daily life—from social media and search engines to healthcare, governance, and education—was rapidly transformed. AI provided answers in seconds, generating texts, images, and even videos that seemed perfect.

Yet, this perfection blurred the line between real and fake, human and machine. The efficiency and flawlessness of AI made us question what truly moves us—the creations of algorithms or the authenticity of human life.

Zhihu posed its annual question: “2025, what is real?” inviting people to reflect and respond through everyday experiences. Moments like a father reading in his final days, a mother slowly recovering from illness, a street vendor’s humble but heartfelt service, or Nobel laureate Yang Zhenning’s mentorship—all reveal truths AI cannot replicate.

Second Me’s digital avatars demonstrate that AI can store memory, but it cannot live, feel, or make choices like humans. Reality does not belong to flawless algorithms—it belongs to those who engage with the world through action and emotion.

In a year dominated by AI, we realized history is written through individual human experiences. Life itself is the ultimate answer.

接著讀

ByteDance Unveils GR-3: A General-Purpose Robot “Brain” That Learns New Tasks with Minimal Data and Masters Flexible Objects

On July 22, ByteDance’s Seed team introduced GR-3, a next-generation Vision-Language-Action (VLA) model. Unlike previous systems that require massive robot trajectory datasets, GR-3 can be efficiently fine-tuned on new tasks or unseen objects with just a handful of human demonstrations. It understands abstract language commands and performs dexterous manipulation of soft objects.

415 天前