The Best AI Models in the World Can’t Pretend to Be Human. That Is a Bigger Problem Than It Sounds.
中文摘要
世界顶尖AI无法模仿人类对话，这比听起来更严重。
English Summary
Top AI models fail human conversation simulation, a significant problem.
Original Excerpt
A new benchmark tested every major frontier model on simulating how real people talk. GPT-6 Astra scored zero. So did Claude Fable 5. Continue reading on Medium »