Neural networks have learned to pretend to be people so well that even other AI owners can't understand them Scientists tested ChatGPT, Claude, Gemini and Grok: out of 98 attempts to recognize the AI, they succeeded only three times. And when the bots were given human biographies and taught to make typos, the score dropped to zero. Ironically, neural networkers have often been considered bots rather than bots themselves. You can't trust anyone anymore.

Neural networks have learned to pretend to be people so well that even other AI owners can't understand them Scientists tested ChatGPT, Claude, Gemini and Grok: out of 98 attempts to recognize the AI, they succeeded only three times. And when the bots were given human biographies and taught to make typos, the score dropped to zero. Ironically, neural networkers have often been considered bots rather than bots themselves. You can't trust anyone anymore.