戳破AI万能论，带你看清真相
Gary Marcus 在 X 上引用 METR_Evals 的结果，反驳某英国大报声称 AI 能完成任何人类任务的论断。他指出 Google 连可靠计数都做不到，且 METR 的测试仅针对软件领域，不要求完全准确。许多人类数秒即可完成的任务，AI 仍然无法可靠执行。
I read this garbage (in a big UK newspaper) when Google can’t even count reliably and wonder why peo...
I read this garbage (in a big UK newspaper) when Google can’t even count reliably and wonder why people don’t spend more time learning about AI’s actual strengths and weakness before running their mouths. the @METR_Evals results alluded to below absolutely do not show that AI can complete any human task of a given length. they pertain to software, and do not demand full accuracy. Many human tasks, even short ones can be reliably be performed by humans in seconds or minutes yet still elude AI. 💬 0 🔄 0 ❤️ 7 👀 1275 📊 1 ⚡