Back to NewsRead Original
AI News CN (Telegram) - English Translation

AI Encounters a Soul-searching Question! This Question Causes All Models to Fail Miserably

AI Faces a Soul-searching Question! This Problem Makes All Models Fail Collectively

via cnBeta.COM Chinese Industry News Site - Telegram Channel

Telegraph
AI Faces a Soul-searching Question! This Problem Makes All Models Fail Collectively

AI Faces a Soul-searching Question! This Problem Makes All Models Fail Collectively. In response to this difficult problem, large models at home and abroad that support image input have given different answers. Among them, o3 gave an answer of 45, and Gemini 2.5Pro gave an answer of only 10. For domestic large models, such as DeepSeek and Qwen3, the answers are 14 and 9 respectively. What is the reason for these different answers? Please read on. Why are there different answers? The core reason: Large models have different understandings of the specifications of the large cube formed. o3 understands the specifications of the finally formed large cube as 5x5x5, but it still gives the wrong answer to the number of missing small cubes. From a human visual perspective, to form a 5x5x5 large cube requires 125…

•••